1 -----------------------------------------------------------------------------
2 This file contains a concatenation of the PCRE man pages, converted to plain
3 text format for ease of searching with a text editor, or for use on systems
4 that do not have a man page processor. The small individual files that give
5 synopses of each function in the library have not been included. Neither has
6 the pcredemo program. There are separate text files for the pcregrep and
8 -----------------------------------------------------------------------------
11 PCRE(3) Library Functions Manual PCRE(3)
16 PCRE - Perl-compatible regular expressions
20 The PCRE library is a set of functions that implement regular expres-
21 sion pattern matching using the same syntax and semantics as Perl, with
22 just a few differences. Some features that appeared in Python and PCRE
23 before they appeared in Perl are also available using the Python syn-
24 tax, there is some support for one or two .NET and Oniguruma syntax
25 items, and there is an option for requesting some minor changes that
26 give better JavaScript compatibility.
28 Starting with release 8.30, it is possible to compile two separate PCRE
29 libraries: the original, which supports 8-bit character strings
30 (including UTF-8 strings), and a second library that supports 16-bit
31 character strings (including UTF-16 strings). The build process allows
32 either one or both to be built. The majority of the work to make this
33 possible was done by Zoltan Herczeg.
35 Starting with release 8.32 it is possible to compile a third separate
36 PCRE library that supports 32-bit character strings (including UTF-32
37 strings). The build process allows any combination of the 8-, 16- and
38 32-bit libraries. The work to make this possible was done by Christian
41 The three libraries contain identical sets of functions, except that
42 the names in the 16-bit library start with pcre16_ instead of pcre_,
43 and the names in the 32-bit library start with pcre32_ instead of
44 pcre_. To avoid over-complication and reduce the documentation mainte-
45 nance load, most of the documentation describes the 8-bit library, with
46 the differences for the 16-bit and 32-bit libraries described sepa-
47 rately in the pcre16 and pcre32 pages. References to functions or
48 structures of the form pcre[16|32]_xxx should be read as meaning
49 "pcre_xxx when using the 8-bit library, pcre16_xxx when using the
50 16-bit library, or pcre32_xxx when using the 32-bit library".
52 The current implementation of PCRE corresponds approximately with Perl
53 5.12, including support for UTF-8/16/32 encoded strings and Unicode
54 general category properties. However, UTF-8/16/32 and Unicode support
55 has to be explicitly enabled; it is not the default. The Unicode tables
56 correspond to Unicode release 6.3.0.
58 In addition to the Perl-compatible matching function, PCRE contains an
59 alternative function that matches the same compiled patterns in a dif-
60 ferent way. In certain circumstances, the alternative function has some
61 advantages. For a discussion of the two matching algorithms, see the
64 PCRE is written in C and released as a C library. A number of people
65 have written wrappers and interfaces of various kinds. In particular,
66 Google Inc. have provided a comprehensive C++ wrapper for the 8-bit
67 library. This is now included as part of the PCRE distribution. The
68 pcrecpp page has details of this interface. Other people's contribu-
69 tions can be found in the Contrib directory at the primary FTP site,
72 ftp://ftp.csx.cam.ac.uk/pub/software/programming/pcre
74 Details of exactly which Perl regular expression features are and are
75 not supported by PCRE are given in separate documents. See the pcrepat-
76 tern and pcrecompat pages. There is a syntax summary in the pcresyntax
79 Some features of PCRE can be included, excluded, or changed when the
80 library is built. The pcre_config() function makes it possible for a
81 client to discover which features are available. The features them-
82 selves are described in the pcrebuild page. Documentation about build-
83 ing PCRE for various operating systems can be found in the README and
84 NON-AUTOTOOLS_BUILD files in the source distribution.
86 The libraries contains a number of undocumented internal functions and
87 data tables that are used by more than one of the exported external
88 functions, but which are not intended for use by external callers.
89 Their names all begin with "_pcre_" or "_pcre16_" or "_pcre32_", which
90 hopefully will not provoke any name clashes. In some environments, it
91 is possible to control which external symbols are exported when a
92 shared library is built, and in these cases the undocumented symbols
96 SECURITY CONSIDERATIONS
98 If you are using PCRE in a non-UTF application that permits users to
99 supply arbitrary patterns for compilation, you should be aware of a
100 feature that allows users to turn on UTF support from within a pattern,
101 provided that PCRE was built with UTF support. For example, an 8-bit
102 pattern that begins with "(*UTF8)" or "(*UTF)" turns on UTF-8 mode,
103 which interprets patterns and subjects as strings of UTF-8 characters
104 instead of individual 8-bit characters. This causes both the pattern
105 and any data against which it is matched to be checked for UTF-8 valid-
106 ity. If the data string is very long, such a check might use suffi-
107 ciently many resources as to cause your application to lose perfor-
110 One way of guarding against this possibility is to use the
111 pcre_fullinfo() function to check the compiled pattern's options for
112 UTF. Alternatively, from release 8.33, you can set the PCRE_NEVER_UTF
113 option at compile time. This causes an compile time error if a pattern
114 contains a UTF-setting sequence.
116 If your application is one that supports UTF, be aware that validity
117 checking can take time. If the same data string is to be matched many
118 times, you can use the PCRE_NO_UTF[8|16|32]_CHECK option for the second
119 and subsequent matches to save redundant checks.
121 Another way that performance can be hit is by running a pattern that
122 has a very large search tree against a string that will never match.
123 Nested unlimited repeats in a pattern are a common example. PCRE pro-
124 vides some protection against this: see the PCRE_EXTRA_MATCH_LIMIT fea-
125 ture in the pcreapi page.
130 The user documentation for PCRE comprises a number of different sec-
131 tions. In the "man" format, each of these is a separate "man page". In
132 the HTML format, each is a separate page, linked from the index page.
133 In the plain text format, the descriptions of the pcregrep and pcretest
134 programs are in files called pcregrep.txt and pcretest.txt, respec-
135 tively. The remaining sections, except for the pcredemo section (which
136 is a program listing), are concatenated in pcre.txt, for ease of
137 searching. The sections are as follows:
140 pcre-config show PCRE installation configuration information
141 pcre16 details of the 16-bit library
142 pcre32 details of the 32-bit library
143 pcreapi details of PCRE's native C API
144 pcrebuild building PCRE
145 pcrecallout details of the callout feature
146 pcrecompat discussion of Perl compatibility
147 pcrecpp details of the C++ wrapper for the 8-bit library
148 pcredemo a demonstration C program that uses PCRE
149 pcregrep description of the pcregrep command (8-bit only)
150 pcrejit discussion of the just-in-time optimization support
151 pcrelimits details of size and other limits
152 pcrematching discussion of the two matching algorithms
153 pcrepartial details of the partial matching facility
154 pcrepattern syntax and semantics of supported
156 pcreperform discussion of performance issues
157 pcreposix the POSIX-compatible C API for the 8-bit library
158 pcreprecompile details of saving and re-using precompiled patterns
159 pcresample discussion of the pcredemo program
160 pcrestack discussion of stack usage
161 pcresyntax quick syntax reference
162 pcretest description of the pcretest testing command
163 pcreunicode discussion of Unicode and UTF-8/16/32 support
165 In the "man" and HTML formats, there is also a short page for each C
166 library function, listing its arguments and results.
172 University Computing Service
173 Cambridge CB2 3QH, England.
175 Putting an actual email address here seems to have been a spam magnet,
176 so I've taken it away. If you want to email me, use my two initials,
177 followed by the two digits 10, at the domain cam.ac.uk.
182 Last updated: 08 January 2014
183 Copyright (c) 1997-2014 University of Cambridge.
184 ------------------------------------------------------------------------------
187 PCRE(3) Library Functions Manual PCRE(3)
192 PCRE - Perl-compatible regular expressions
197 PCRE 16-BIT API BASIC FUNCTIONS
199 pcre16 *pcre16_compile(PCRE_SPTR16 pattern, int options,
200 const char **errptr, int *erroffset,
201 const unsigned char *tableptr);
203 pcre16 *pcre16_compile2(PCRE_SPTR16 pattern, int options,
205 const char **errptr, int *erroffset,
206 const unsigned char *tableptr);
208 pcre16_extra *pcre16_study(const pcre16 *code, int options,
209 const char **errptr);
211 void pcre16_free_study(pcre16_extra *extra);
213 int pcre16_exec(const pcre16 *code, const pcre16_extra *extra,
214 PCRE_SPTR16 subject, int length, int startoffset,
215 int options, int *ovector, int ovecsize);
217 int pcre16_dfa_exec(const pcre16 *code, const pcre16_extra *extra,
218 PCRE_SPTR16 subject, int length, int startoffset,
219 int options, int *ovector, int ovecsize,
220 int *workspace, int wscount);
223 PCRE 16-BIT API STRING EXTRACTION FUNCTIONS
225 int pcre16_copy_named_substring(const pcre16 *code,
226 PCRE_SPTR16 subject, int *ovector,
227 int stringcount, PCRE_SPTR16 stringname,
228 PCRE_UCHAR16 *buffer, int buffersize);
230 int pcre16_copy_substring(PCRE_SPTR16 subject, int *ovector,
231 int stringcount, int stringnumber, PCRE_UCHAR16 *buffer,
234 int pcre16_get_named_substring(const pcre16 *code,
235 PCRE_SPTR16 subject, int *ovector,
236 int stringcount, PCRE_SPTR16 stringname,
237 PCRE_SPTR16 *stringptr);
239 int pcre16_get_stringnumber(const pcre16 *code,
242 int pcre16_get_stringtable_entries(const pcre16 *code,
243 PCRE_SPTR16 name, PCRE_UCHAR16 **first, PCRE_UCHAR16 **last);
245 int pcre16_get_substring(PCRE_SPTR16 subject, int *ovector,
246 int stringcount, int stringnumber,
247 PCRE_SPTR16 *stringptr);
249 int pcre16_get_substring_list(PCRE_SPTR16 subject,
250 int *ovector, int stringcount, PCRE_SPTR16 **listptr);
252 void pcre16_free_substring(PCRE_SPTR16 stringptr);
254 void pcre16_free_substring_list(PCRE_SPTR16 *stringptr);
257 PCRE 16-BIT API AUXILIARY FUNCTIONS
259 pcre16_jit_stack *pcre16_jit_stack_alloc(int startsize, int maxsize);
261 void pcre16_jit_stack_free(pcre16_jit_stack *stack);
263 void pcre16_assign_jit_stack(pcre16_extra *extra,
264 pcre16_jit_callback callback, void *data);
266 const unsigned char *pcre16_maketables(void);
268 int pcre16_fullinfo(const pcre16 *code, const pcre16_extra *extra,
269 int what, void *where);
271 int pcre16_refcount(pcre16 *code, int adjust);
273 int pcre16_config(int what, void *where);
275 const char *pcre16_version(void);
277 int pcre16_pattern_to_host_byte_order(pcre16 *code,
278 pcre16_extra *extra, const unsigned char *tables);
281 PCRE 16-BIT API INDIRECTED FUNCTIONS
283 void *(*pcre16_malloc)(size_t);
285 void (*pcre16_free)(void *);
287 void *(*pcre16_stack_malloc)(size_t);
289 void (*pcre16_stack_free)(void *);
291 int (*pcre16_callout)(pcre16_callout_block *);
294 PCRE 16-BIT API 16-BIT-ONLY FUNCTION
296 int pcre16_utf16_to_host_byte_order(PCRE_UCHAR16 *output,
297 PCRE_SPTR16 input, int length, int *byte_order,
301 THE PCRE 16-BIT LIBRARY
303 Starting with release 8.30, it is possible to compile a PCRE library
304 that supports 16-bit character strings, including UTF-16 strings, as
305 well as or instead of the original 8-bit library. The majority of the
306 work to make this possible was done by Zoltan Herczeg. The two
307 libraries contain identical sets of functions, used in exactly the same
308 way. Only the names of the functions and the data types of their argu-
309 ments and results are different. To avoid over-complication and reduce
310 the documentation maintenance load, most of the PCRE documentation
311 describes the 8-bit library, with only occasional references to the
312 16-bit library. This page describes what is different when you use the
315 WARNING: A single application can be linked with both libraries, but
316 you must take care when processing any particular pattern to use func-
317 tions from just one library. For example, if you want to study a pat-
318 tern that was compiled with pcre16_compile(), you must do so with
319 pcre16_study(), not pcre_study(), and you must free the study data with
325 There is only one header file, pcre.h. It contains prototypes for all
326 the functions in all libraries, as well as definitions of flags, struc-
327 tures, error codes, etc.
332 In Unix-like systems, the 16-bit library is called libpcre16, and can
333 normally be accesss by adding -lpcre16 to the command for linking an
334 application that uses PCRE.
339 In the 8-bit library, strings are passed to PCRE library functions as
340 vectors of bytes with the C type "char *". In the 16-bit library,
341 strings are passed as vectors of unsigned 16-bit quantities. The macro
342 PCRE_UCHAR16 specifies an appropriate data type, and PCRE_SPTR16 is
343 defined as "const PCRE_UCHAR16 *". In very many environments, "short
344 int" is a 16-bit data type. When PCRE is built, it defines PCRE_UCHAR16
345 as "unsigned short int", but checks that it really is a 16-bit data
346 type. If it is not, the build fails with an error message telling the
347 maintainer to modify the definition appropriately.
352 The types of the opaque structures that are used for compiled 16-bit
353 patterns and JIT stacks are pcre16 and pcre16_jit_stack respectively.
354 The type of the user-accessible structure that is returned by
355 pcre16_study() is pcre16_extra, and the type of the structure that is
356 used for passing data to a callout function is pcre16_callout_block.
357 These structures contain the same fields, with the same names, as their
358 8-bit counterparts. The only difference is that pointers to character
359 strings are 16-bit instead of 8-bit types.
364 For every function in the 8-bit library there is a corresponding func-
365 tion in the 16-bit library with a name that starts with pcre16_ instead
366 of pcre_. The prototypes are listed above. In addition, there is one
367 extra function, pcre16_utf16_to_host_byte_order(). This is a utility
368 function that converts a UTF-16 character string to host byte order if
369 necessary. The other 16-bit functions expect the strings they are
370 passed to be in host byte order.
372 The input and output arguments of pcre16_utf16_to_host_byte_order() may
373 point to the same address, that is, conversion in place is supported.
374 The output buffer must be at least as long as the input.
376 The length argument specifies the number of 16-bit data units in the
377 input string; a negative value specifies a zero-terminated string.
379 If byte_order is NULL, it is assumed that the string starts off in host
380 byte order. This may be changed by byte-order marks (BOMs) anywhere in
381 the string (commonly as the first character).
383 If byte_order is not NULL, a non-zero value of the integer to which it
384 points means that the input starts off in host byte order, otherwise
385 the opposite order is assumed. Again, BOMs in the string can change
386 this. The final byte order is passed back at the end of processing.
388 If keep_boms is not zero, byte-order mark characters (0xfeff) are
389 copied into the output string. Otherwise they are discarded.
391 The result of the function is the number of 16-bit units placed into
392 the output buffer, including the zero terminator if the string was
396 SUBJECT STRING OFFSETS
398 The lengths and starting offsets of subject strings must be specified
399 in 16-bit data units, and the offsets within subject strings that are
400 returned by the matching functions are in also 16-bit units rather than
406 The name-to-number translation table that is maintained for named sub-
407 patterns uses 16-bit characters. The pcre16_get_stringtable_entries()
408 function returns the length of each entry in the table as the number of
414 There are two new general option names, PCRE_UTF16 and
415 PCRE_NO_UTF16_CHECK, which correspond to PCRE_UTF8 and
416 PCRE_NO_UTF8_CHECK in the 8-bit library. In fact, these new options
417 define the same bits in the options word. There is a discussion about
418 the validity of UTF-16 strings in the pcreunicode page.
420 For the pcre16_config() function there is an option PCRE_CONFIG_UTF16
421 that returns 1 if UTF-16 support is configured, otherwise 0. If this
422 option is given to pcre_config() or pcre32_config(), or if the
423 PCRE_CONFIG_UTF8 or PCRE_CONFIG_UTF32 option is given to pcre16_con-
424 fig(), the result is the PCRE_ERROR_BADOPTION error.
429 In 16-bit mode, when PCRE_UTF16 is not set, character values are
430 treated in the same way as in 8-bit, non UTF-8 mode, except, of course,
431 that they can range from 0 to 0xffff instead of 0 to 0xff. Character
432 types for characters less than 0xff can therefore be influenced by the
433 locale in the same way as before. Characters greater than 0xff have
434 only one case, and no "type" (such as letter or digit).
436 In UTF-16 mode, the character code is Unicode, in the range 0 to
437 0x10ffff, with the exception of values in the range 0xd800 to 0xdfff
438 because those are "surrogate" values that are used in pairs to encode
439 values greater than 0xffff.
441 A UTF-16 string can indicate its endianness by special code knows as a
442 byte-order mark (BOM). The PCRE functions do not handle this, expecting
443 strings to be in host byte order. A utility function called
444 pcre16_utf16_to_host_byte_order() is provided to help with this (see
450 The errors PCRE_ERROR_BADUTF16_OFFSET and PCRE_ERROR_SHORTUTF16 corre-
451 spond to their 8-bit counterparts. The error PCRE_ERROR_BADMODE is
452 given when a compiled pattern is passed to a function that processes
453 patterns in the other mode, for example, if a pattern compiled with
454 pcre_compile() is passed to pcre16_exec().
456 There are new error codes whose names begin with PCRE_UTF16_ERR for
457 invalid UTF-16 strings, corresponding to the PCRE_UTF8_ERR codes for
458 UTF-8 strings that are described in the section entitled "Reason codes
459 for invalid UTF-8 strings" in the main pcreapi page. The UTF-16 errors
462 PCRE_UTF16_ERR1 Missing low surrogate at end of string
463 PCRE_UTF16_ERR2 Invalid low surrogate follows high surrogate
464 PCRE_UTF16_ERR3 Isolated low surrogate
465 PCRE_UTF16_ERR4 Non-character
470 If there is an error while compiling a pattern, the error text that is
471 passed back by pcre16_compile() or pcre16_compile2() is still an 8-bit
472 character string, zero-terminated.
477 The subject and mark fields in the callout block that is passed to a
478 callout function point to 16-bit vectors.
483 The pcretest program continues to operate with 8-bit input and output
484 files, but it can be used for testing the 16-bit library. If it is run
485 with the command line option -16, patterns and subject strings are con-
486 verted from 8-bit to 16-bit before being passed to PCRE, and the 16-bit
487 library functions are used instead of the 8-bit ones. Returned 16-bit
488 strings are converted to 8-bit for output. If both the 8-bit and the
489 32-bit libraries were not compiled, pcretest defaults to 16-bit and the
490 -16 option is ignored.
492 When PCRE is being built, the RunTest script that is called by "make
493 check" uses the pcretest -C option to discover which of the 8-bit,
494 16-bit and 32-bit libraries has been built, and runs the tests appro-
498 NOT SUPPORTED IN 16-BIT MODE
500 Not all the features of the 8-bit library are available with the 16-bit
501 library. The C++ and POSIX wrapper functions support only the 8-bit
502 library, and the pcregrep program is at present 8-bit only.
508 University Computing Service
509 Cambridge CB2 3QH, England.
514 Last updated: 12 May 2013
515 Copyright (c) 1997-2013 University of Cambridge.
516 ------------------------------------------------------------------------------
519 PCRE(3) Library Functions Manual PCRE(3)
524 PCRE - Perl-compatible regular expressions
529 PCRE 32-BIT API BASIC FUNCTIONS
531 pcre32 *pcre32_compile(PCRE_SPTR32 pattern, int options,
532 const char **errptr, int *erroffset,
533 const unsigned char *tableptr);
535 pcre32 *pcre32_compile2(PCRE_SPTR32 pattern, int options,
537 const unsigned char *tableptr);
539 pcre32_extra *pcre32_study(const pcre32 *code, int options,
540 const char **errptr);
542 void pcre32_free_study(pcre32_extra *extra);
544 int pcre32_exec(const pcre32 *code, const pcre32_extra *extra,
545 PCRE_SPTR32 subject, int length, int startoffset,
546 int options, int *ovector, int ovecsize);
548 int pcre32_dfa_exec(const pcre32 *code, const pcre32_extra *extra,
549 PCRE_SPTR32 subject, int length, int startoffset,
550 int options, int *ovector, int ovecsize,
551 int *workspace, int wscount);
554 PCRE 32-BIT API STRING EXTRACTION FUNCTIONS
556 int pcre32_copy_named_substring(const pcre32 *code,
557 PCRE_SPTR32 subject, int *ovector,
558 int stringcount, PCRE_SPTR32 stringname,
559 PCRE_UCHAR32 *buffer, int buffersize);
561 int pcre32_copy_substring(PCRE_SPTR32 subject, int *ovector,
562 int stringcount, int stringnumber, PCRE_UCHAR32 *buffer,
565 int pcre32_get_named_substring(const pcre32 *code,
566 PCRE_SPTR32 subject, int *ovector,
567 int stringcount, PCRE_SPTR32 stringname,
568 PCRE_SPTR32 *stringptr);
570 int pcre32_get_stringnumber(const pcre32 *code,
573 int pcre32_get_stringtable_entries(const pcre32 *code,
574 PCRE_SPTR32 name, PCRE_UCHAR32 **first, PCRE_UCHAR32 **last);
576 int pcre32_get_substring(PCRE_SPTR32 subject, int *ovector,
577 int stringcount, int stringnumber,
578 PCRE_SPTR32 *stringptr);
580 int pcre32_get_substring_list(PCRE_SPTR32 subject,
581 int *ovector, int stringcount, PCRE_SPTR32 **listptr);
583 void pcre32_free_substring(PCRE_SPTR32 stringptr);
585 void pcre32_free_substring_list(PCRE_SPTR32 *stringptr);
588 PCRE 32-BIT API AUXILIARY FUNCTIONS
590 pcre32_jit_stack *pcre32_jit_stack_alloc(int startsize, int maxsize);
592 void pcre32_jit_stack_free(pcre32_jit_stack *stack);
594 void pcre32_assign_jit_stack(pcre32_extra *extra,
595 pcre32_jit_callback callback, void *data);
597 const unsigned char *pcre32_maketables(void);
599 int pcre32_fullinfo(const pcre32 *code, const pcre32_extra *extra,
600 int what, void *where);
602 int pcre32_refcount(pcre32 *code, int adjust);
604 int pcre32_config(int what, void *where);
606 const char *pcre32_version(void);
608 int pcre32_pattern_to_host_byte_order(pcre32 *code,
609 pcre32_extra *extra, const unsigned char *tables);
612 PCRE 32-BIT API INDIRECTED FUNCTIONS
614 void *(*pcre32_malloc)(size_t);
616 void (*pcre32_free)(void *);
618 void *(*pcre32_stack_malloc)(size_t);
620 void (*pcre32_stack_free)(void *);
622 int (*pcre32_callout)(pcre32_callout_block *);
625 PCRE 32-BIT API 32-BIT-ONLY FUNCTION
627 int pcre32_utf32_to_host_byte_order(PCRE_UCHAR32 *output,
628 PCRE_SPTR32 input, int length, int *byte_order,
632 THE PCRE 32-BIT LIBRARY
634 Starting with release 8.32, it is possible to compile a PCRE library
635 that supports 32-bit character strings, including UTF-32 strings, as
636 well as or instead of the original 8-bit library. This work was done by
637 Christian Persch, based on the work done by Zoltan Herczeg for the
638 16-bit library. All three libraries contain identical sets of func-
639 tions, used in exactly the same way. Only the names of the functions
640 and the data types of their arguments and results are different. To
641 avoid over-complication and reduce the documentation maintenance load,
642 most of the PCRE documentation describes the 8-bit library, with only
643 occasional references to the 16-bit and 32-bit libraries. This page
644 describes what is different when you use the 32-bit library.
646 WARNING: A single application can be linked with all or any of the
647 three libraries, but you must take care when processing any particular
648 pattern to use functions from just one library. For example, if you
649 want to study a pattern that was compiled with pcre32_compile(), you
650 must do so with pcre32_study(), not pcre_study(), and you must free the
651 study data with pcre32_free_study().
656 There is only one header file, pcre.h. It contains prototypes for all
657 the functions in all libraries, as well as definitions of flags, struc-
658 tures, error codes, etc.
663 In Unix-like systems, the 32-bit library is called libpcre32, and can
664 normally be accesss by adding -lpcre32 to the command for linking an
665 application that uses PCRE.
670 In the 8-bit library, strings are passed to PCRE library functions as
671 vectors of bytes with the C type "char *". In the 32-bit library,
672 strings are passed as vectors of unsigned 32-bit quantities. The macro
673 PCRE_UCHAR32 specifies an appropriate data type, and PCRE_SPTR32 is
674 defined as "const PCRE_UCHAR32 *". In very many environments, "unsigned
675 int" is a 32-bit data type. When PCRE is built, it defines PCRE_UCHAR32
676 as "unsigned int", but checks that it really is a 32-bit data type. If
677 it is not, the build fails with an error message telling the maintainer
678 to modify the definition appropriately.
683 The types of the opaque structures that are used for compiled 32-bit
684 patterns and JIT stacks are pcre32 and pcre32_jit_stack respectively.
685 The type of the user-accessible structure that is returned by
686 pcre32_study() is pcre32_extra, and the type of the structure that is
687 used for passing data to a callout function is pcre32_callout_block.
688 These structures contain the same fields, with the same names, as their
689 8-bit counterparts. The only difference is that pointers to character
690 strings are 32-bit instead of 8-bit types.
695 For every function in the 8-bit library there is a corresponding func-
696 tion in the 32-bit library with a name that starts with pcre32_ instead
697 of pcre_. The prototypes are listed above. In addition, there is one
698 extra function, pcre32_utf32_to_host_byte_order(). This is a utility
699 function that converts a UTF-32 character string to host byte order if
700 necessary. The other 32-bit functions expect the strings they are
701 passed to be in host byte order.
703 The input and output arguments of pcre32_utf32_to_host_byte_order() may
704 point to the same address, that is, conversion in place is supported.
705 The output buffer must be at least as long as the input.
707 The length argument specifies the number of 32-bit data units in the
708 input string; a negative value specifies a zero-terminated string.
710 If byte_order is NULL, it is assumed that the string starts off in host
711 byte order. This may be changed by byte-order marks (BOMs) anywhere in
712 the string (commonly as the first character).
714 If byte_order is not NULL, a non-zero value of the integer to which it
715 points means that the input starts off in host byte order, otherwise
716 the opposite order is assumed. Again, BOMs in the string can change
717 this. The final byte order is passed back at the end of processing.
719 If keep_boms is not zero, byte-order mark characters (0xfeff) are
720 copied into the output string. Otherwise they are discarded.
722 The result of the function is the number of 32-bit units placed into
723 the output buffer, including the zero terminator if the string was
727 SUBJECT STRING OFFSETS
729 The lengths and starting offsets of subject strings must be specified
730 in 32-bit data units, and the offsets within subject strings that are
731 returned by the matching functions are in also 32-bit units rather than
737 The name-to-number translation table that is maintained for named sub-
738 patterns uses 32-bit characters. The pcre32_get_stringtable_entries()
739 function returns the length of each entry in the table as the number of
745 There are two new general option names, PCRE_UTF32 and
746 PCRE_NO_UTF32_CHECK, which correspond to PCRE_UTF8 and
747 PCRE_NO_UTF8_CHECK in the 8-bit library. In fact, these new options
748 define the same bits in the options word. There is a discussion about
749 the validity of UTF-32 strings in the pcreunicode page.
751 For the pcre32_config() function there is an option PCRE_CONFIG_UTF32
752 that returns 1 if UTF-32 support is configured, otherwise 0. If this
753 option is given to pcre_config() or pcre16_config(), or if the
754 PCRE_CONFIG_UTF8 or PCRE_CONFIG_UTF16 option is given to pcre32_con-
755 fig(), the result is the PCRE_ERROR_BADOPTION error.
760 In 32-bit mode, when PCRE_UTF32 is not set, character values are
761 treated in the same way as in 8-bit, non UTF-8 mode, except, of course,
762 that they can range from 0 to 0x7fffffff instead of 0 to 0xff. Charac-
763 ter types for characters less than 0xff can therefore be influenced by
764 the locale in the same way as before. Characters greater than 0xff
765 have only one case, and no "type" (such as letter or digit).
767 In UTF-32 mode, the character code is Unicode, in the range 0 to
768 0x10ffff, with the exception of values in the range 0xd800 to 0xdfff
769 because those are "surrogate" values that are ill-formed in UTF-32.
771 A UTF-32 string can indicate its endianness by special code knows as a
772 byte-order mark (BOM). The PCRE functions do not handle this, expecting
773 strings to be in host byte order. A utility function called
774 pcre32_utf32_to_host_byte_order() is provided to help with this (see
780 The error PCRE_ERROR_BADUTF32 corresponds to its 8-bit counterpart.
781 The error PCRE_ERROR_BADMODE is given when a compiled pattern is passed
782 to a function that processes patterns in the other mode, for example,
783 if a pattern compiled with pcre_compile() is passed to pcre32_exec().
785 There are new error codes whose names begin with PCRE_UTF32_ERR for
786 invalid UTF-32 strings, corresponding to the PCRE_UTF8_ERR codes for
787 UTF-8 strings that are described in the section entitled "Reason codes
788 for invalid UTF-8 strings" in the main pcreapi page. The UTF-32 errors
791 PCRE_UTF32_ERR1 Surrogate character (range from 0xd800 to 0xdfff)
792 PCRE_UTF32_ERR2 Non-character
793 PCRE_UTF32_ERR3 Character > 0x10ffff
798 If there is an error while compiling a pattern, the error text that is
799 passed back by pcre32_compile() or pcre32_compile2() is still an 8-bit
800 character string, zero-terminated.
805 The subject and mark fields in the callout block that is passed to a
806 callout function point to 32-bit vectors.
811 The pcretest program continues to operate with 8-bit input and output
812 files, but it can be used for testing the 32-bit library. If it is run
813 with the command line option -32, patterns and subject strings are con-
814 verted from 8-bit to 32-bit before being passed to PCRE, and the 32-bit
815 library functions are used instead of the 8-bit ones. Returned 32-bit
816 strings are converted to 8-bit for output. If both the 8-bit and the
817 16-bit libraries were not compiled, pcretest defaults to 32-bit and the
818 -32 option is ignored.
820 When PCRE is being built, the RunTest script that is called by "make
821 check" uses the pcretest -C option to discover which of the 8-bit,
822 16-bit and 32-bit libraries has been built, and runs the tests appro-
826 NOT SUPPORTED IN 32-BIT MODE
828 Not all the features of the 8-bit library are available with the 32-bit
829 library. The C++ and POSIX wrapper functions support only the 8-bit
830 library, and the pcregrep program is at present 8-bit only.
836 University Computing Service
837 Cambridge CB2 3QH, England.
842 Last updated: 12 May 2013
843 Copyright (c) 1997-2013 University of Cambridge.
844 ------------------------------------------------------------------------------
847 PCREBUILD(3) Library Functions Manual PCREBUILD(3)
852 PCRE - Perl-compatible regular expressions
856 PCRE is distributed with a configure script that can be used to build
857 the library in Unix-like environments using the applications known as
858 Autotools. Also in the distribution are files to support building
859 using CMake instead of configure. The text file README contains general
860 information about building with Autotools (some of which is repeated
861 below), and also has some comments about building on various operating
862 systems. There is a lot more information about building PCRE without
863 using Autotools (including information about using CMake and building
864 "by hand") in the text file called NON-AUTOTOOLS-BUILD. You should
865 consult this file as well as the README file if you are building in a
866 non-Unix-like environment.
869 PCRE BUILD-TIME OPTIONS
871 The rest of this document describes the optional features of PCRE that
872 can be selected when the library is compiled. It assumes use of the
873 configure script, where the optional features are selected or dese-
874 lected by providing options to configure before running the make com-
875 mand. However, the same options can be selected in both Unix-like and
876 non-Unix-like environments using the GUI facility of cmake-gui if you
877 are using CMake instead of configure to build PCRE.
879 If you are not using Autotools or CMake, option selection can be done
880 by editing the config.h file, or by passing parameter settings to the
881 compiler, as described in NON-AUTOTOOLS-BUILD.
883 The complete list of options for configure (which includes the standard
884 ones such as the selection of the installation directory) can be
889 The following sections include descriptions of options whose names
890 begin with --enable or --disable. These settings specify changes to the
891 defaults for the configure command. Because of the way that configure
892 works, --enable and --disable always come in pairs, so the complemen-
893 tary option always exists as well, but as it specifies the default, it
897 BUILDING 8-BIT, 16-BIT AND 32-BIT LIBRARIES
899 By default, a library called libpcre is built, containing functions
900 that take string arguments contained in vectors of bytes, either as
901 single-byte characters, or interpreted as UTF-8 strings. You can also
902 build a separate library, called libpcre16, in which strings are con-
903 tained in vectors of 16-bit data units and interpreted either as sin-
904 gle-unit characters or UTF-16 strings, by adding
908 to the configure command. You can also build yet another separate
909 library, called libpcre32, in which strings are contained in vectors of
910 32-bit data units and interpreted either as single-unit characters or
911 UTF-32 strings, by adding
915 to the configure command. If you do not want the 8-bit library, add
919 as well. At least one of the three libraries must be built. Note that
920 the C++ and POSIX wrappers are for the 8-bit library only, and that
921 pcregrep is an 8-bit program. None of these are built if you select
922 only the 16-bit or 32-bit libraries.
925 BUILDING SHARED AND STATIC LIBRARIES
927 The Autotools PCRE building process uses libtool to build both shared
928 and static libraries by default. You can suppress one of these by
934 to the configure command, as required.
939 By default, if the 8-bit library is being built, the configure script
940 will search for a C++ compiler and C++ header files. If it finds them,
941 it automatically builds the C++ wrapper library (which supports only
942 8-bit strings). You can disable this by adding
946 to the configure command.
949 UTF-8, UTF-16 AND UTF-32 SUPPORT
951 To build PCRE with support for UTF Unicode character strings, add
955 to the configure command. This setting applies to all three libraries,
956 adding support for UTF-8 to the 8-bit library, support for UTF-16 to
957 the 16-bit library, and support for UTF-32 to the to the 32-bit
958 library. There are no separate options for enabling UTF-8, UTF-16 and
959 UTF-32 independently because that would allow ridiculous settings such
960 as requesting UTF-16 support while building only the 8-bit library. It
961 is not possible to build one library with UTF support and another with-
962 out in the same configuration. (For backwards compatibility, --enable-
963 utf8 is a synonym of --enable-utf.)
965 Of itself, this setting does not make PCRE treat strings as UTF-8,
966 UTF-16 or UTF-32. As well as compiling PCRE with this option, you also
967 have have to set the PCRE_UTF8, PCRE_UTF16 or PCRE_UTF32 option (as
968 appropriate) when you call one of the pattern compiling functions.
970 If you set --enable-utf when compiling in an EBCDIC environment, PCRE
971 expects its input to be either ASCII or UTF-8 (depending on the run-
972 time option). It is not possible to support both EBCDIC and UTF-8 codes
973 in the same version of the library. Consequently, --enable-utf and
974 --enable-ebcdic are mutually exclusive.
977 UNICODE CHARACTER PROPERTY SUPPORT
979 UTF support allows the libraries to process character codepoints up to
980 0x10ffff in the strings that they handle. On its own, however, it does
981 not provide any facilities for accessing the properties of such charac-
982 ters. If you want to be able to use the pattern escapes \P, \p, and \X,
983 which refer to Unicode character properties, you must add
985 --enable-unicode-properties
987 to the configure command. This implies UTF support, even if you have
988 not explicitly requested it.
990 Including Unicode property support adds around 30K of tables to the
991 PCRE library. Only the general category properties such as Lu and Nd
992 are supported. Details are given in the pcrepattern documentation.
995 JUST-IN-TIME COMPILER SUPPORT
997 Just-in-time compiler support is included in the build by specifying
1001 This support is available only for certain hardware architectures. If
1002 this option is set for an unsupported architecture, a compile time
1003 error occurs. See the pcrejit documentation for a discussion of JIT
1004 usage. When JIT support is enabled, pcregrep automatically makes use of
1007 --disable-pcregrep-jit
1009 to the "configure" command.
1012 CODE VALUE OF NEWLINE
1014 By default, PCRE interprets the linefeed (LF) character as indicating
1015 the end of a line. This is the normal newline character on Unix-like
1016 systems. You can compile PCRE to use carriage return (CR) instead, by
1019 --enable-newline-is-cr
1021 to the configure command. There is also a --enable-newline-is-lf
1022 option, which explicitly specifies linefeed as the newline character.
1024 Alternatively, you can specify that line endings are to be indicated by
1025 the two character sequence CRLF. If you want this, add
1027 --enable-newline-is-crlf
1029 to the configure command. There is a fourth option, specified by
1031 --enable-newline-is-anycrlf
1033 which causes PCRE to recognize any of the three sequences CR, LF, or
1034 CRLF as indicating a line ending. Finally, a fifth option, specified by
1036 --enable-newline-is-any
1038 causes PCRE to recognize any Unicode newline sequence.
1040 Whatever line ending convention is selected when PCRE is built can be
1041 overridden when the library functions are called. At build time it is
1042 conventional to use the standard for your operating system.
1047 By default, the sequence \R in a pattern matches any Unicode newline
1048 sequence, whatever has been selected as the line ending sequence. If
1051 --enable-bsr-anycrlf
1053 the default is changed so that \R matches only CR, LF, or CRLF. What-
1054 ever is selected when PCRE is built can be overridden when the library
1055 functions are called.
1060 When the 8-bit library is called through the POSIX interface (see the
1061 pcreposix documentation), additional working storage is required for
1062 holding the pointers to capturing substrings, because PCRE requires
1063 three integers per substring, whereas the POSIX interface provides only
1064 two. If the number of expected substrings is small, the wrapper func-
1065 tion uses space on the stack, because this is faster than using mal-
1066 loc() for each call. The default threshold above which the stack is no
1067 longer used is 10; it can be changed by adding a setting such as
1069 --with-posix-malloc-threshold=20
1071 to the configure command.
1074 HANDLING VERY LARGE PATTERNS
1076 Within a compiled pattern, offset values are used to point from one
1077 part to another (for example, from an opening parenthesis to an alter-
1078 nation metacharacter). By default, in the 8-bit and 16-bit libraries,
1079 two-byte values are used for these offsets, leading to a maximum size
1080 for a compiled pattern of around 64K. This is sufficient to handle all
1081 but the most gigantic patterns. Nevertheless, some people do want to
1082 process truly enormous patterns, so it is possible to compile PCRE to
1083 use three-byte or four-byte offsets by adding a setting such as
1087 to the configure command. The value given must be 2, 3, or 4. For the
1088 16-bit library, a value of 3 is rounded up to 4. In these libraries,
1089 using longer offsets slows down the operation of PCRE because it has to
1090 load additional data when handling them. For the 32-bit library the
1091 value is always 4 and cannot be overridden; the value of --with-link-
1095 AVOIDING EXCESSIVE STACK USAGE
1097 When matching with the pcre_exec() function, PCRE implements backtrack-
1098 ing by making recursive calls to an internal function called match().
1099 In environments where the size of the stack is limited, this can se-
1100 verely limit PCRE's operation. (The Unix environment does not usually
1101 suffer from this problem, but it may sometimes be necessary to increase
1102 the maximum stack size. There is a discussion in the pcrestack docu-
1103 mentation.) An alternative approach to recursion that uses memory from
1104 the heap to remember data, instead of using recursive function calls,
1105 has been implemented to work round the problem of limited stack size.
1106 If you want to build a version of PCRE that works this way, add
1108 --disable-stack-for-recursion
1110 to the configure command. With this configuration, PCRE will use the
1111 pcre_stack_malloc and pcre_stack_free variables to call memory manage-
1112 ment functions. By default these point to malloc() and free(), but you
1113 can replace the pointers so that your own functions are used instead.
1115 Separate functions are provided rather than using pcre_malloc and
1116 pcre_free because the usage is very predictable: the block sizes
1117 requested are always the same, and the blocks are always freed in
1118 reverse order. A calling program might be able to implement optimized
1119 functions that perform better than malloc() and free(). PCRE runs
1120 noticeably more slowly when built in this way. This option affects only
1121 the pcre_exec() function; it is not relevant for pcre_dfa_exec().
1124 LIMITING PCRE RESOURCE USAGE
1126 Internally, PCRE has a function called match(), which it calls repeat-
1127 edly (sometimes recursively) when matching a pattern with the
1128 pcre_exec() function. By controlling the maximum number of times this
1129 function may be called during a single matching operation, a limit can
1130 be placed on the resources used by a single call to pcre_exec(). The
1131 limit can be changed at run time, as described in the pcreapi documen-
1132 tation. The default is 10 million, but this can be changed by adding a
1135 --with-match-limit=500000
1137 to the configure command. This setting has no effect on the
1138 pcre_dfa_exec() matching function.
1140 In some environments it is desirable to limit the depth of recursive
1141 calls of match() more strictly than the total number of calls, in order
1142 to restrict the maximum amount of stack (or heap, if --disable-stack-
1143 for-recursion is specified) that is used. A second limit controls this;
1144 it defaults to the value that is set for --with-match-limit, which
1145 imposes no additional constraints. However, you can set a lower limit
1146 by adding, for example,
1148 --with-match-limit-recursion=10000
1150 to the configure command. This value can also be overridden at run
1154 CREATING CHARACTER TABLES AT BUILD TIME
1156 PCRE uses fixed tables for processing characters whose code values are
1157 less than 256. By default, PCRE is built with a set of tables that are
1158 distributed in the file pcre_chartables.c.dist. These tables are for
1159 ASCII codes only. If you add
1161 --enable-rebuild-chartables
1163 to the configure command, the distributed tables are no longer used.
1164 Instead, a program called dftables is compiled and run. This outputs
1165 the source for new set of tables, created in the default locale of your
1166 C run-time system. (This method of replacing the tables does not work
1167 if you are cross compiling, because dftables is run on the local host.
1168 If you need to create alternative tables when cross compiling, you will
1169 have to do so "by hand".)
1174 PCRE assumes by default that it will run in an environment where the
1175 character code is ASCII (or Unicode, which is a superset of ASCII).
1176 This is the case for most computer operating systems. PCRE can, how-
1177 ever, be compiled to run in an EBCDIC environment by adding
1181 to the configure command. This setting implies --enable-rebuild-charta-
1182 bles. You should only use it if you know that you are in an EBCDIC
1183 environment (for example, an IBM mainframe operating system). The
1184 --enable-ebcdic option is incompatible with --enable-utf.
1186 The EBCDIC character that corresponds to an ASCII LF is assumed to have
1187 the value 0x15 by default. However, in some EBCDIC environments, 0x25
1188 is used. In such an environment you should use
1190 --enable-ebcdic-nl25
1192 as well as, or instead of, --enable-ebcdic. The EBCDIC character for CR
1193 has the same value as in ASCII, namely, 0x0d. Whichever of 0x15 and
1194 0x25 is not chosen as LF is made to correspond to the Unicode NEL char-
1195 acter (which, in Unicode, is 0x85).
1197 The options that select newline behaviour, such as --enable-newline-is-
1198 cr, and equivalent run-time options, refer to these character values in
1199 an EBCDIC environment.
1202 PCREGREP OPTIONS FOR COMPRESSED FILE SUPPORT
1204 By default, pcregrep reads all files as plain text. You can build it so
1205 that it recognizes files whose names end in .gz or .bz2, and reads them
1206 with libz or libbz2, respectively, by adding one or both of
1208 --enable-pcregrep-libz
1209 --enable-pcregrep-libbz2
1211 to the configure command. These options naturally require that the rel-
1212 evant libraries are installed on your system. Configuration will fail
1216 PCREGREP BUFFER SIZE
1218 pcregrep uses an internal buffer to hold a "window" on the file it is
1219 scanning, in order to be able to output "before" and "after" lines when
1220 it finds a match. The size of the buffer is controlled by a parameter
1221 whose default value is 20K. The buffer itself is three times this size,
1222 but because of the way it is used for holding "before" lines, the long-
1223 est line that is guaranteed to be processable is the parameter size.
1224 You can change the default parameter value by adding, for example,
1226 --with-pcregrep-bufsize=50K
1228 to the configure command. The caller of pcregrep can, however, override
1229 this value by specifying a run-time option.
1232 PCRETEST OPTION FOR LIBREADLINE SUPPORT
1236 --enable-pcretest-libreadline
1238 to the configure command, pcretest is linked with the libreadline
1239 library, and when its input is from a terminal, it reads it using the
1240 readline() function. This provides line-editing and history facilities.
1241 Note that libreadline is GPL-licensed, so if you distribute a binary of
1242 pcretest linked in this way, there may be licensing issues.
1244 Setting this option causes the -lreadline option to be added to the
1245 pcretest build. In many operating environments with a sytem-installed
1246 libreadline this is sufficient. However, in some environments (e.g. if
1247 an unmodified distribution version of readline is in use), some extra
1248 configuration may be necessary. The INSTALL file for libreadline says
1251 "Readline uses the termcap functions, but does not link with the
1252 termcap or curses library itself, allowing applications which link
1253 with readline the to choose an appropriate library."
1255 If your environment has not been set up so that an appropriate library
1256 is automatically included, you may need to add something like
1260 immediately before the configure command.
1263 DEBUGGING WITH VALGRIND SUPPORT
1269 option to to the configure command, PCRE will use valgrind annotations
1270 to mark certain memory regions as unaddressable. This allows it to
1271 detect invalid memory accesses, and is mostly useful for debugging PCRE
1275 CODE COVERAGE REPORTING
1277 If your C compiler is gcc, you can build a version of PCRE that can
1278 generate a code coverage report for its test suite. To enable this, you
1279 must install lcov version 1.6 or above. Then specify
1283 to the configure command and build PCRE in the usual way.
1285 Note that using ccache (a caching C compiler) is incompatible with code
1286 coverage reporting. If you have configured ccache to run automatically
1287 on your system, you must set the environment variable
1291 before running make to build PCRE, so that ccache is not used.
1293 When --enable-coverage is used, the following addition targets are
1294 added to the Makefile:
1298 This creates a fresh coverage report for the PCRE test suite. It is
1299 equivalent to running "make coverage-reset", "make coverage-baseline",
1300 "make check", and then "make coverage-report".
1304 This zeroes the coverage counters, but does nothing else.
1306 make coverage-baseline
1308 This captures baseline coverage information.
1310 make coverage-report
1312 This creates the coverage report.
1314 make coverage-clean-report
1316 This removes the generated coverage report without cleaning the cover-
1319 make coverage-clean-data
1321 This removes the captured coverage data without removing the coverage
1322 files created at compile time (*.gcno).
1326 This cleans all coverage data including the generated coverage report.
1327 For more information about code coverage, see the gcov and lcov docu-
1333 pcreapi(3), pcre16, pcre32, pcre_config(3).
1339 University Computing Service
1340 Cambridge CB2 3QH, England.
1345 Last updated: 12 May 2013
1346 Copyright (c) 1997-2013 University of Cambridge.
1347 ------------------------------------------------------------------------------
1350 PCREMATCHING(3) Library Functions Manual PCREMATCHING(3)
1355 PCRE - Perl-compatible regular expressions
1357 PCRE MATCHING ALGORITHMS
1359 This document describes the two different algorithms that are available
1360 in PCRE for matching a compiled regular expression against a given sub-
1361 ject string. The "standard" algorithm is the one provided by the
1362 pcre_exec(), pcre16_exec() and pcre32_exec() functions. These work in
1363 the same as as Perl's matching function, and provide a Perl-compatible
1364 matching operation. The just-in-time (JIT) optimization that is
1365 described in the pcrejit documentation is compatible with these func-
1368 An alternative algorithm is provided by the pcre_dfa_exec(),
1369 pcre16_dfa_exec() and pcre32_dfa_exec() functions; they operate in a
1370 different way, and are not Perl-compatible. This alternative has advan-
1371 tages and disadvantages compared with the standard algorithm, and these
1372 are described below.
1374 When there is only one possible way in which a given subject string can
1375 match a pattern, the two algorithms give the same answer. A difference
1376 arises, however, when there are multiple possibilities. For example, if
1381 is matched against the string
1383 <something> <something else> <something further>
1385 there are three possible answers. The standard algorithm finds only one
1386 of them, whereas the alternative algorithm finds all three.
1389 REGULAR EXPRESSIONS AS TREES
1391 The set of strings that are matched by a regular expression can be rep-
1392 resented as a tree structure. An unlimited repetition in the pattern
1393 makes the tree of infinite size, but it is still a tree. Matching the
1394 pattern to a given subject string (from a given starting point) can be
1395 thought of as a search of the tree. There are two ways to search a
1396 tree: depth-first and breadth-first, and these correspond to the two
1397 matching algorithms provided by PCRE.
1400 THE STANDARD MATCHING ALGORITHM
1402 In the terminology of Jeffrey Friedl's book "Mastering Regular Expres-
1403 sions", the standard algorithm is an "NFA algorithm". It conducts a
1404 depth-first search of the pattern tree. That is, it proceeds along a
1405 single path through the tree, checking that the subject matches what is
1406 required. When there is a mismatch, the algorithm tries any alterna-
1407 tives at the current point, and if they all fail, it backs up to the
1408 previous branch point in the tree, and tries the next alternative
1409 branch at that level. This often involves backing up (moving to the
1410 left) in the subject string as well. The order in which repetition
1411 branches are tried is controlled by the greedy or ungreedy nature of
1414 If a leaf node is reached, a matching string has been found, and at
1415 that point the algorithm stops. Thus, if there is more than one possi-
1416 ble match, this algorithm returns the first one that it finds. Whether
1417 this is the shortest, the longest, or some intermediate length depends
1418 on the way the greedy and ungreedy repetition quantifiers are specified
1421 Because it ends up with a single path through the tree, it is rela-
1422 tively straightforward for this algorithm to keep track of the sub-
1423 strings that are matched by portions of the pattern in parentheses.
1424 This provides support for capturing parentheses and back references.
1427 THE ALTERNATIVE MATCHING ALGORITHM
1429 This algorithm conducts a breadth-first search of the tree. Starting
1430 from the first matching point in the subject, it scans the subject
1431 string from left to right, once, character by character, and as it does
1432 this, it remembers all the paths through the tree that represent valid
1433 matches. In Friedl's terminology, this is a kind of "DFA algorithm",
1434 though it is not implemented as a traditional finite state machine (it
1435 keeps multiple states active simultaneously).
1437 Although the general principle of this matching algorithm is that it
1438 scans the subject string only once, without backtracking, there is one
1439 exception: when a lookaround assertion is encountered, the characters
1440 following or preceding the current point have to be independently
1443 The scan continues until either the end of the subject is reached, or
1444 there are no more unterminated paths. At this point, terminated paths
1445 represent the different matching possibilities (if there are none, the
1446 match has failed). Thus, if there is more than one possible match,
1447 this algorithm finds all of them, and in particular, it finds the long-
1448 est. The matches are returned in decreasing order of length. There is
1449 an option to stop the algorithm after the first match (which is neces-
1450 sarily the shortest) is found.
1452 Note that all the matches that are found start at the same point in the
1453 subject. If the pattern
1457 is matched against the string "the caterpillar catchment", the result
1458 will be the three strings "caterpillar", "cater", and "cat" that start
1459 at the fifth character of the subject. The algorithm does not automati-
1460 cally move on to find matches that start at later positions.
1462 PCRE's "auto-possessification" optimization usually applies to charac-
1463 ter repeats at the end of a pattern (as well as internally). For exam-
1464 ple, the pattern "a\d+" is compiled as if it were "a\d++" because there
1465 is no point even considering the possibility of backtracking into the
1466 repeated digits. For DFA matching, this means that only one possible
1467 match is found. If you really do want multiple matches in such cases,
1468 either use an ungreedy repeat ("a\d+?") or set the PCRE_NO_AUTO_POSSESS
1469 option when compiling.
1471 There are a number of features of PCRE regular expressions that are not
1472 supported by the alternative matching algorithm. They are as follows:
1474 1. Because the algorithm finds all possible matches, the greedy or
1475 ungreedy nature of repetition quantifiers is not relevant. Greedy and
1476 ungreedy quantifiers are treated in exactly the same way. However, pos-
1477 sessive quantifiers can make a difference when what follows could also
1478 match what is quantified, for example in a pattern like this:
1482 This pattern matches "aaab!" but not "aaa!", which would be matched by
1483 a non-possessive quantifier. Similarly, if an atomic group is present,
1484 it is matched as if it were a standalone pattern at the current point,
1485 and the longest match is then "locked in" for the rest of the overall
1488 2. When dealing with multiple paths through the tree simultaneously, it
1489 is not straightforward to keep track of captured substrings for the
1490 different matching possibilities, and PCRE's implementation of this
1491 algorithm does not attempt to do this. This means that no captured sub-
1492 strings are available.
1494 3. Because no substrings are captured, back references within the pat-
1495 tern are not supported, and cause errors if encountered.
1497 4. For the same reason, conditional expressions that use a backrefer-
1498 ence as the condition or test for a specific group recursion are not
1501 5. Because many paths through the tree may be active, the \K escape
1502 sequence, which resets the start of the match when encountered (but may
1503 be on some paths and not on others), is not supported. It causes an
1504 error if encountered.
1506 6. Callouts are supported, but the value of the capture_top field is
1507 always 1, and the value of the capture_last field is always -1.
1509 7. The \C escape sequence, which (in the standard algorithm) always
1510 matches a single data unit, even in UTF-8, UTF-16 or UTF-32 modes, is
1511 not supported in these modes, because the alternative algorithm moves
1512 through the subject string one character (not data unit) at a time, for
1513 all active paths through the tree.
1515 8. Except for (*FAIL), the backtracking control verbs such as (*PRUNE)
1516 are not supported. (*FAIL) is supported, and behaves like a failing
1520 ADVANTAGES OF THE ALTERNATIVE ALGORITHM
1522 Using the alternative matching algorithm provides the following advan-
1525 1. All possible matches (at a single point in the subject) are automat-
1526 ically found, and in particular, the longest match is found. To find
1527 more than one match using the standard algorithm, you have to do kludgy
1528 things with callouts.
1530 2. Because the alternative algorithm scans the subject string just
1531 once, and never needs to backtrack (except for lookbehinds), it is pos-
1532 sible to pass very long subject strings to the matching function in
1533 several pieces, checking for partial matching each time. Although it is
1534 possible to do multi-segment matching using the standard algorithm by
1535 retaining partially matched substrings, it is more complicated. The
1536 pcrepartial documentation gives details of partial matching and dis-
1537 cusses multi-segment matching.
1540 DISADVANTAGES OF THE ALTERNATIVE ALGORITHM
1542 The alternative algorithm suffers from a number of disadvantages:
1544 1. It is substantially slower than the standard algorithm. This is
1545 partly because it has to search for all possible matches, but is also
1546 because it is less susceptible to optimization.
1548 2. Capturing parentheses and back references are not supported.
1550 3. Although atomic groups are supported, their use does not provide the
1551 performance advantage that it does for the standard algorithm.
1557 University Computing Service
1558 Cambridge CB2 3QH, England.
1563 Last updated: 12 November 2013
1564 Copyright (c) 1997-2012 University of Cambridge.
1565 ------------------------------------------------------------------------------
1568 PCREAPI(3) Library Functions Manual PCREAPI(3)
1573 PCRE - Perl-compatible regular expressions
1578 PCRE NATIVE API BASIC FUNCTIONS
1580 pcre *pcre_compile(const char *pattern, int options,
1581 const char **errptr, int *erroffset,
1582 const unsigned char *tableptr);
1584 pcre *pcre_compile2(const char *pattern, int options,
1586 const char **errptr, int *erroffset,
1587 const unsigned char *tableptr);
1589 pcre_extra *pcre_study(const pcre *code, int options,
1590 const char **errptr);
1592 void pcre_free_study(pcre_extra *extra);
1594 int pcre_exec(const pcre *code, const pcre_extra *extra,
1595 const char *subject, int length, int startoffset,
1596 int options, int *ovector, int ovecsize);
1598 int pcre_dfa_exec(const pcre *code, const pcre_extra *extra,
1599 const char *subject, int length, int startoffset,
1600 int options, int *ovector, int ovecsize,
1601 int *workspace, int wscount);
1604 PCRE NATIVE API STRING EXTRACTION FUNCTIONS
1606 int pcre_copy_named_substring(const pcre *code,
1607 const char *subject, int *ovector,
1608 int stringcount, const char *stringname,
1609 char *buffer, int buffersize);
1611 int pcre_copy_substring(const char *subject, int *ovector,
1612 int stringcount, int stringnumber, char *buffer,
1615 int pcre_get_named_substring(const pcre *code,
1616 const char *subject, int *ovector,
1617 int stringcount, const char *stringname,
1618 const char **stringptr);
1620 int pcre_get_stringnumber(const pcre *code,
1623 int pcre_get_stringtable_entries(const pcre *code,
1624 const char *name, char **first, char **last);
1626 int pcre_get_substring(const char *subject, int *ovector,
1627 int stringcount, int stringnumber,
1628 const char **stringptr);
1630 int pcre_get_substring_list(const char *subject,
1631 int *ovector, int stringcount, const char ***listptr);
1633 void pcre_free_substring(const char *stringptr);
1635 void pcre_free_substring_list(const char **stringptr);
1638 PCRE NATIVE API AUXILIARY FUNCTIONS
1640 int pcre_jit_exec(const pcre *code, const pcre_extra *extra,
1641 const char *subject, int length, int startoffset,
1642 int options, int *ovector, int ovecsize,
1643 pcre_jit_stack *jstack);
1645 pcre_jit_stack *pcre_jit_stack_alloc(int startsize, int maxsize);
1647 void pcre_jit_stack_free(pcre_jit_stack *stack);
1649 void pcre_assign_jit_stack(pcre_extra *extra,
1650 pcre_jit_callback callback, void *data);
1652 const unsigned char *pcre_maketables(void);
1654 int pcre_fullinfo(const pcre *code, const pcre_extra *extra,
1655 int what, void *where);
1657 int pcre_refcount(pcre *code, int adjust);
1659 int pcre_config(int what, void *where);
1661 const char *pcre_version(void);
1663 int pcre_pattern_to_host_byte_order(pcre *code,
1664 pcre_extra *extra, const unsigned char *tables);
1667 PCRE NATIVE API INDIRECTED FUNCTIONS
1669 void *(*pcre_malloc)(size_t);
1671 void (*pcre_free)(void *);
1673 void *(*pcre_stack_malloc)(size_t);
1675 void (*pcre_stack_free)(void *);
1677 int (*pcre_callout)(pcre_callout_block *);
1679 int (*pcre_stack_guard)(void);
1682 PCRE 8-BIT, 16-BIT, AND 32-BIT LIBRARIES
1684 As well as support for 8-bit character strings, PCRE also supports
1685 16-bit strings (from release 8.30) and 32-bit strings (from release
1686 8.32), by means of two additional libraries. They can be built as well
1687 as, or instead of, the 8-bit library. To avoid too much complication,
1688 this document describes the 8-bit versions of the functions, with only
1689 occasional references to the 16-bit and 32-bit libraries.
1691 The 16-bit and 32-bit functions operate in the same way as their 8-bit
1692 counterparts; they just use different data types for their arguments
1693 and results, and their names start with pcre16_ or pcre32_ instead of
1694 pcre_. For every option that has UTF8 in its name (for example,
1695 PCRE_UTF8), there are corresponding 16-bit and 32-bit names with UTF8
1696 replaced by UTF16 or UTF32, respectively. This facility is in fact just
1697 cosmetic; the 16-bit and 32-bit option names define the same bit val-
1700 References to bytes and UTF-8 in this document should be read as refer-
1701 ences to 16-bit data units and UTF-16 when using the 16-bit library, or
1702 32-bit data units and UTF-32 when using the 32-bit library, unless
1703 specified otherwise. More details of the specific differences for the
1704 16-bit and 32-bit libraries are given in the pcre16 and pcre32 pages.
1709 PCRE has its own native API, which is described in this document. There
1710 are also some wrapper functions (for the 8-bit library only) that cor-
1711 respond to the POSIX regular expression API, but they do not give
1712 access to all the functionality. They are described in the pcreposix
1713 documentation. Both of these APIs define a set of C function calls. A
1714 C++ wrapper (again for the 8-bit library only) is also distributed with
1715 PCRE. It is documented in the pcrecpp page.
1717 The native API C function prototypes are defined in the header file
1718 pcre.h, and on Unix-like systems the (8-bit) library itself is called
1719 libpcre. It can normally be accessed by adding -lpcre to the command
1720 for linking an application that uses PCRE. The header file defines the
1721 macros PCRE_MAJOR and PCRE_MINOR to contain the major and minor release
1722 numbers for the library. Applications can use these to include support
1723 for different releases of PCRE.
1725 In a Windows environment, if you want to statically link an application
1726 program against a non-dll pcre.a file, you must define PCRE_STATIC
1727 before including pcre.h or pcrecpp.h, because otherwise the pcre_mal-
1728 loc() and pcre_free() exported functions will be declared
1729 __declspec(dllimport), with unwanted results.
1731 The functions pcre_compile(), pcre_compile2(), pcre_study(), and
1732 pcre_exec() are used for compiling and matching regular expressions in
1733 a Perl-compatible manner. A sample program that demonstrates the sim-
1734 plest way of using them is provided in the file called pcredemo.c in
1735 the PCRE source distribution. A listing of this program is given in the
1736 pcredemo documentation, and the pcresample documentation describes how
1737 to compile and run it.
1739 Just-in-time compiler support is an optional feature of PCRE that can
1740 be built in appropriate hardware environments. It greatly speeds up the
1741 matching performance of many patterns. Simple programs can easily
1742 request that it be used if available, by setting an option that is
1743 ignored when it is not relevant. More complicated programs might need
1744 to make use of the functions pcre_jit_stack_alloc(),
1745 pcre_jit_stack_free(), and pcre_assign_jit_stack() in order to control
1746 the JIT code's memory usage.
1748 From release 8.32 there is also a direct interface for JIT execution,
1749 which gives improved performance. The JIT-specific functions are dis-
1750 cussed in the pcrejit documentation.
1752 A second matching function, pcre_dfa_exec(), which is not Perl-compati-
1753 ble, is also provided. This uses a different algorithm for the match-
1754 ing. The alternative algorithm finds all possible matches (at a given
1755 point in the subject), and scans the subject just once (unless there
1756 are lookbehind assertions). However, this algorithm does not return
1757 captured substrings. A description of the two matching algorithms and
1758 their advantages and disadvantages is given in the pcrematching docu-
1761 In addition to the main compiling and matching functions, there are
1762 convenience functions for extracting captured substrings from a subject
1763 string that is matched by pcre_exec(). They are:
1765 pcre_copy_substring()
1766 pcre_copy_named_substring()
1767 pcre_get_substring()
1768 pcre_get_named_substring()
1769 pcre_get_substring_list()
1770 pcre_get_stringnumber()
1771 pcre_get_stringtable_entries()
1773 pcre_free_substring() and pcre_free_substring_list() are also provided,
1774 to free the memory used for extracted strings.
1776 The function pcre_maketables() is used to build a set of character
1777 tables in the current locale for passing to pcre_compile(),
1778 pcre_exec(), or pcre_dfa_exec(). This is an optional facility that is
1779 provided for specialist use. Most commonly, no special tables are
1780 passed, in which case internal tables that are generated when PCRE is
1783 The function pcre_fullinfo() is used to find out information about a
1784 compiled pattern. The function pcre_version() returns a pointer to a
1785 string containing the version of PCRE and its date of release.
1787 The function pcre_refcount() maintains a reference count in a data
1788 block containing a compiled pattern. This is provided for the benefit
1789 of object-oriented applications.
1791 The global variables pcre_malloc and pcre_free initially contain the
1792 entry points of the standard malloc() and free() functions, respec-
1793 tively. PCRE calls the memory management functions via these variables,
1794 so a calling program can replace them if it wishes to intercept the
1795 calls. This should be done before calling any PCRE functions.
1797 The global variables pcre_stack_malloc and pcre_stack_free are also
1798 indirections to memory management functions. These special functions
1799 are used only when PCRE is compiled to use the heap for remembering
1800 data, instead of recursive function calls, when running the pcre_exec()
1801 function. See the pcrebuild documentation for details of how to do
1802 this. It is a non-standard way of building PCRE, for use in environ-
1803 ments that have limited stacks. Because of the greater use of memory
1804 management, it runs more slowly. Separate functions are provided so
1805 that special-purpose external code can be used for this case. When
1806 used, these functions are always called in a stack-like manner (last
1807 obtained, first freed), and always for memory blocks of the same size.
1808 There is a discussion about PCRE's stack usage in the pcrestack docu-
1811 The global variable pcre_callout initially contains NULL. It can be set
1812 by the caller to a "callout" function, which PCRE will then call at
1813 specified points during a matching operation. Details are given in the
1814 pcrecallout documentation.
1816 The global variable pcre_stack_guard initially contains NULL. It can be
1817 set by the caller to a function that is called by PCRE whenever it
1818 starts to compile a parenthesized part of a pattern. When parentheses
1819 are nested, PCRE uses recursive function calls, which use up the system
1820 stack. This function is provided so that applications with restricted
1821 stacks can force a compilation error if the stack runs out. The func-
1822 tion should return zero if all is well, or non-zero to force an error.
1827 PCRE supports five different conventions for indicating line breaks in
1828 strings: a single CR (carriage return) character, a single LF (line-
1829 feed) character, the two-character sequence CRLF, any of the three pre-
1830 ceding, or any Unicode newline sequence. The Unicode newline sequences
1831 are the three just mentioned, plus the single characters VT (vertical
1832 tab, U+000B), FF (form feed, U+000C), NEL (next line, U+0085), LS (line
1833 separator, U+2028), and PS (paragraph separator, U+2029).
1835 Each of the first three conventions is used by at least one operating
1836 system as its standard newline sequence. When PCRE is built, a default
1837 can be specified. The default default is LF, which is the Unix stan-
1838 dard. When PCRE is run, the default can be overridden, either when a
1839 pattern is compiled, or when it is matched.
1841 At compile time, the newline convention can be specified by the options
1842 argument of pcre_compile(), or it can be specified by special text at
1843 the start of the pattern itself; this overrides any other settings. See
1844 the pcrepattern page for details of the special character sequences.
1846 In the PCRE documentation the word "newline" is used to mean "the char-
1847 acter or pair of characters that indicate a line break". The choice of
1848 newline convention affects the handling of the dot, circumflex, and
1849 dollar metacharacters, the handling of #-comments in /x mode, and, when
1850 CRLF is a recognized line ending sequence, the match position advance-
1851 ment for a non-anchored pattern. There is more detail about this in the
1852 section on pcre_exec() options below.
1854 The choice of newline convention does not affect the interpretation of
1855 the \n or \r escape sequences, nor does it affect what \R matches,
1856 which is controlled in a similar way, but by separate options.
1861 The PCRE functions can be used in multi-threading applications, with
1862 the proviso that the memory management functions pointed to by
1863 pcre_malloc, pcre_free, pcre_stack_malloc, and pcre_stack_free, and the
1864 callout and stack-checking functions pointed to by pcre_callout and
1865 pcre_stack_guard, are shared by all threads.
1867 The compiled form of a regular expression is not altered during match-
1868 ing, so the same compiled pattern can safely be used by several threads
1871 If the just-in-time optimization feature is being used, it needs sepa-
1872 rate memory stack areas for each thread. See the pcrejit documentation
1876 SAVING PRECOMPILED PATTERNS FOR LATER USE
1878 The compiled form of a regular expression can be saved and re-used at a
1879 later time, possibly by a different program, and even on a host other
1880 than the one on which it was compiled. Details are given in the
1881 pcreprecompile documentation, which includes a description of the
1882 pcre_pattern_to_host_byte_order() function. However, compiling a regu-
1883 lar expression with one version of PCRE for use with a different ver-
1884 sion is not guaranteed to work and may cause crashes.
1887 CHECKING BUILD-TIME OPTIONS
1889 int pcre_config(int what, void *where);
1891 The function pcre_config() makes it possible for a PCRE client to dis-
1892 cover which optional features have been compiled into the PCRE library.
1893 The pcrebuild documentation has more details about these optional fea-
1896 The first argument for pcre_config() is an integer, specifying which
1897 information is required; the second argument is a pointer to a variable
1898 into which the information is placed. The returned value is zero on
1899 success, or the negative error code PCRE_ERROR_BADOPTION if the value
1900 in the first argument is not recognized. The following information is
1905 The output is an integer that is set to one if UTF-8 support is avail-
1906 able; otherwise it is set to zero. This value should normally be given
1907 to the 8-bit version of this function, pcre_config(). If it is given to
1908 the 16-bit or 32-bit version of this function, the result is
1909 PCRE_ERROR_BADOPTION.
1913 The output is an integer that is set to one if UTF-16 support is avail-
1914 able; otherwise it is set to zero. This value should normally be given
1915 to the 16-bit version of this function, pcre16_config(). If it is given
1916 to the 8-bit or 32-bit version of this function, the result is
1917 PCRE_ERROR_BADOPTION.
1921 The output is an integer that is set to one if UTF-32 support is avail-
1922 able; otherwise it is set to zero. This value should normally be given
1923 to the 32-bit version of this function, pcre32_config(). If it is given
1924 to the 8-bit or 16-bit version of this function, the result is
1925 PCRE_ERROR_BADOPTION.
1927 PCRE_CONFIG_UNICODE_PROPERTIES
1929 The output is an integer that is set to one if support for Unicode
1930 character properties is available; otherwise it is set to zero.
1934 The output is an integer that is set to one if support for just-in-time
1935 compiling is available; otherwise it is set to zero.
1937 PCRE_CONFIG_JITTARGET
1939 The output is a pointer to a zero-terminated "const char *" string. If
1940 JIT support is available, the string contains the name of the architec-
1941 ture for which the JIT compiler is configured, for example "x86 32bit
1942 (little endian + unaligned)". If JIT support is not available, the
1947 The output is an integer whose value specifies the default character
1948 sequence that is recognized as meaning "newline". The values that are
1949 supported in ASCII/Unicode environments are: 10 for LF, 13 for CR, 3338
1950 for CRLF, -2 for ANYCRLF, and -1 for ANY. In EBCDIC environments, CR,
1951 ANYCRLF, and ANY yield the same values. However, the value for LF is
1952 normally 21, though some EBCDIC environments use 37. The corresponding
1953 values for CRLF are 3349 and 3365. The default should normally corre-
1954 spond to the standard sequence for your operating system.
1958 The output is an integer whose value indicates what character sequences
1959 the \R escape sequence matches by default. A value of 0 means that \R
1960 matches any Unicode line ending sequence; a value of 1 means that \R
1961 matches only CR, LF, or CRLF. The default can be overridden when a pat-
1962 tern is compiled or matched.
1964 PCRE_CONFIG_LINK_SIZE
1966 The output is an integer that contains the number of bytes used for
1967 internal linkage in compiled regular expressions. For the 8-bit
1968 library, the value can be 2, 3, or 4. For the 16-bit library, the value
1969 is either 2 or 4 and is still a number of bytes. For the 32-bit
1970 library, the value is either 2 or 4 and is still a number of bytes. The
1971 default value of 2 is sufficient for all but the most massive patterns,
1972 since it allows the compiled pattern to be up to 64K in size. Larger
1973 values allow larger regular expressions to be compiled, at the expense
1976 PCRE_CONFIG_POSIX_MALLOC_THRESHOLD
1978 The output is an integer that contains the threshold above which the
1979 POSIX interface uses malloc() for output vectors. Further details are
1980 given in the pcreposix documentation.
1982 PCRE_CONFIG_PARENS_LIMIT
1984 The output is a long integer that gives the maximum depth of nesting of
1985 parentheses (of any kind) in a pattern. This limit is imposed to cap
1986 the amount of system stack used when a pattern is compiled. It is spec-
1987 ified when PCRE is built; the default is 250. This limit does not take
1988 into account the stack that may already be used by the calling applica-
1989 tion. For finer control over compilation stack usage, you can set a
1990 pointer to an external checking function in pcre_stack_guard.
1992 PCRE_CONFIG_MATCH_LIMIT
1994 The output is a long integer that gives the default limit for the num-
1995 ber of internal matching function calls in a pcre_exec() execution.
1996 Further details are given with pcre_exec() below.
1998 PCRE_CONFIG_MATCH_LIMIT_RECURSION
2000 The output is a long integer that gives the default limit for the depth
2001 of recursion when calling the internal matching function in a
2002 pcre_exec() execution. Further details are given with pcre_exec()
2005 PCRE_CONFIG_STACKRECURSE
2007 The output is an integer that is set to one if internal recursion when
2008 running pcre_exec() is implemented by recursive function calls that use
2009 the stack to remember their state. This is the usual way that PCRE is
2010 compiled. The output is zero if PCRE was compiled to use blocks of data
2011 on the heap instead of recursive function calls. In this case,
2012 pcre_stack_malloc and pcre_stack_free are called to manage memory
2013 blocks on the heap, thus avoiding the use of the stack.
2018 pcre *pcre_compile(const char *pattern, int options,
2019 const char **errptr, int *erroffset,
2020 const unsigned char *tableptr);
2022 pcre *pcre_compile2(const char *pattern, int options,
2024 const char **errptr, int *erroffset,
2025 const unsigned char *tableptr);
2027 Either of the functions pcre_compile() or pcre_compile2() can be called
2028 to compile a pattern into an internal form. The only difference between
2029 the two interfaces is that pcre_compile2() has an additional argument,
2030 errorcodeptr, via which a numerical error code can be returned. To
2031 avoid too much repetition, we refer just to pcre_compile() below, but
2032 the information applies equally to pcre_compile2().
2034 The pattern is a C string terminated by a binary zero, and is passed in
2035 the pattern argument. A pointer to a single block of memory that is
2036 obtained via pcre_malloc is returned. This contains the compiled code
2037 and related data. The pcre type is defined for the returned block; this
2038 is a typedef for a structure whose contents are not externally defined.
2039 It is up to the caller to free the memory (via pcre_free) when it is no
2042 Although the compiled code of a PCRE regex is relocatable, that is, it
2043 does not depend on memory location, the complete pcre data block is not
2044 fully relocatable, because it may contain a copy of the tableptr argu-
2045 ment, which is an address (see below).
2047 The options argument contains various bit settings that affect the com-
2048 pilation. It should be zero if no options are required. The available
2049 options are described below. Some of them (in particular, those that
2050 are compatible with Perl, but some others as well) can also be set and
2051 unset from within the pattern (see the detailed description in the
2052 pcrepattern documentation). For those options that can be different in
2053 different parts of the pattern, the contents of the options argument
2054 specifies their settings at the start of compilation and execution. The
2055 PCRE_ANCHORED, PCRE_BSR_xxx, PCRE_NEWLINE_xxx, PCRE_NO_UTF8_CHECK, and
2056 PCRE_NO_START_OPTIMIZE options can be set at the time of matching as
2057 well as at compile time.
2059 If errptr is NULL, pcre_compile() returns NULL immediately. Otherwise,
2060 if compilation of a pattern fails, pcre_compile() returns NULL, and
2061 sets the variable pointed to by errptr to point to a textual error mes-
2062 sage. This is a static string that is part of the library. You must not
2063 try to free it. Normally, the offset from the start of the pattern to
2064 the data unit that was being processed when the error was discovered is
2065 placed in the variable pointed to by erroffset, which must not be NULL
2066 (if it is, an immediate error is given). However, for an invalid UTF-8
2067 or UTF-16 string, the offset is that of the first data unit of the
2070 Some errors are not detected until the whole pattern has been scanned;
2071 in these cases, the offset passed back is the length of the pattern.
2072 Note that the offset is in data units, not characters, even in a UTF
2073 mode. It may sometimes point into the middle of a UTF-8 or UTF-16 char-
2076 If pcre_compile2() is used instead of pcre_compile(), and the error-
2077 codeptr argument is not NULL, a non-zero error code number is returned
2078 via this argument in the event of an error. This is in addition to the
2079 textual error message. Error codes and messages are listed below.
2081 If the final argument, tableptr, is NULL, PCRE uses a default set of
2082 character tables that are built when PCRE is compiled, using the
2083 default C locale. Otherwise, tableptr must be an address that is the
2084 result of a call to pcre_maketables(). This value is stored with the
2085 compiled pattern, and used again by pcre_exec() and pcre_dfa_exec()
2086 when the pattern is matched. For more discussion, see the section on
2087 locale support below.
2089 This code fragment shows a typical straightforward call to pcre_com-
2096 "^A.*Z", /* the pattern */
2097 0, /* default options */
2098 &error, /* for error message */
2099 &erroffset, /* for error offset */
2100 NULL); /* use default character tables */
2102 The following names for option bits are defined in the pcre.h header
2107 If this bit is set, the pattern is forced to be "anchored", that is, it
2108 is constrained to match only at the first matching point in the string
2109 that is being searched (the "subject string"). This effect can also be
2110 achieved by appropriate constructs in the pattern itself, which is the
2111 only way to do it in Perl.
2115 If this bit is set, pcre_compile() automatically inserts callout items,
2116 all with number 255, before each pattern item. For discussion of the
2117 callout facility, see the pcrecallout documentation.
2122 These options (which are mutually exclusive) control what the \R escape
2123 sequence matches. The choice is either to match only CR, LF, or CRLF,
2124 or to match any Unicode newline sequence. The default is specified when
2125 PCRE is built. It can be overridden from within the pattern, or by set-
2126 ting an option when a compiled pattern is matched.
2130 If this bit is set, letters in the pattern match both upper and lower
2131 case letters. It is equivalent to Perl's /i option, and it can be
2132 changed within a pattern by a (?i) option setting. In UTF-8 mode, PCRE
2133 always understands the concept of case for characters whose values are
2134 less than 128, so caseless matching is always possible. For characters
2135 with higher values, the concept of case is supported if PCRE is com-
2136 piled with Unicode property support, but not otherwise. If you want to
2137 use caseless matching for characters 128 and above, you must ensure
2138 that PCRE is compiled with Unicode property support as well as with
2143 If this bit is set, a dollar metacharacter in the pattern matches only
2144 at the end of the subject string. Without this option, a dollar also
2145 matches immediately before a newline at the end of the string (but not
2146 before any other newlines). The PCRE_DOLLAR_ENDONLY option is ignored
2147 if PCRE_MULTILINE is set. There is no equivalent to this option in
2148 Perl, and no way to set it within a pattern.
2152 If this bit is set, a dot metacharacter in the pattern matches a char-
2153 acter of any value, including one that indicates a newline. However, it
2154 only ever matches one character, even if newlines are coded as CRLF.
2155 Without this option, a dot does not match when the current position is
2156 at a newline. This option is equivalent to Perl's /s option, and it can
2157 be changed within a pattern by a (?s) option setting. A negative class
2158 such as [^a] always matches newline characters, independent of the set-
2159 ting of this option.
2163 If this bit is set, names used to identify capturing subpatterns need
2164 not be unique. This can be helpful for certain types of pattern when it
2165 is known that only one instance of the named subpattern can ever be
2166 matched. There are more details of named subpatterns below; see also
2167 the pcrepattern documentation.
2171 If this bit is set, most white space characters in the pattern are
2172 totally ignored except when escaped or inside a character class. How-
2173 ever, white space is not allowed within sequences such as (?> that
2174 introduce various parenthesized subpatterns, nor within a numerical
2175 quantifier such as {1,3}. However, ignorable white space is permitted
2176 between an item and a following quantifier and between a quantifier and
2177 a following + that indicates possessiveness.
2179 White space did not used to include the VT character (code 11), because
2180 Perl did not treat this character as white space. However, Perl changed
2181 at release 5.18, so PCRE followed at release 8.34, and VT is now
2182 treated as white space.
2184 PCRE_EXTENDED also causes characters between an unescaped # outside a
2185 character class and the next newline, inclusive, to be ignored.
2186 PCRE_EXTENDED is equivalent to Perl's /x option, and it can be changed
2187 within a pattern by a (?x) option setting.
2189 Which characters are interpreted as newlines is controlled by the
2190 options passed to pcre_compile() or by a special sequence at the start
2191 of the pattern, as described in the section entitled "Newline conven-
2192 tions" in the pcrepattern documentation. Note that the end of this type
2193 of comment is a literal newline sequence in the pattern; escape
2194 sequences that happen to represent a newline do not count.
2196 This option makes it possible to include comments inside complicated
2197 patterns. Note, however, that this applies only to data characters.
2198 White space characters may never appear within special character
2199 sequences in a pattern, for example within the sequence (?( that intro-
2200 duces a conditional subpattern.
2204 This option was invented in order to turn on additional functionality
2205 of PCRE that is incompatible with Perl, but it is currently of very
2206 little use. When set, any backslash in a pattern that is followed by a
2207 letter that has no special meaning causes an error, thus reserving
2208 these combinations for future expansion. By default, as in Perl, a
2209 backslash followed by a letter with no special meaning is treated as a
2210 literal. (Perl can, however, be persuaded to give an error for this, by
2211 running it with the -w option.) There are at present no other features
2212 controlled by this option. It can also be set by a (?X) option setting
2217 If this option is set, an unanchored pattern is required to match
2218 before or at the first newline in the subject string, though the
2219 matched text may continue over the newline.
2221 PCRE_JAVASCRIPT_COMPAT
2223 If this option is set, PCRE's behaviour is changed in some ways so that
2224 it is compatible with JavaScript rather than Perl. The changes are as
2227 (1) A lone closing square bracket in a pattern causes a compile-time
2228 error, because this is illegal in JavaScript (by default it is treated
2229 as a data character). Thus, the pattern AB]CD becomes illegal when this
2232 (2) At run time, a back reference to an unset subpattern group matches
2233 an empty string (by default this causes the current matching alterna-
2234 tive to fail). A pattern such as (\1)(a) succeeds when this option is
2235 set (assuming it can find an "a" in the subject), whereas it fails by
2236 default, for Perl compatibility.
2238 (3) \U matches an upper case "U" character; by default \U causes a com-
2239 pile time error (Perl uses \U to upper case subsequent characters).
2241 (4) \u matches a lower case "u" character unless it is followed by four
2242 hexadecimal digits, in which case the hexadecimal number defines the
2243 code point to match. By default, \u causes a compile time error (Perl
2244 uses it to upper case the following character).
2246 (5) \x matches a lower case "x" character unless it is followed by two
2247 hexadecimal digits, in which case the hexadecimal number defines the
2248 code point to match. By default, as in Perl, a hexadecimal number is
2249 always expected after \x, but it may have zero, one, or two digits (so,
2250 for example, \xz matches a binary zero character followed by z).
2254 By default, for the purposes of matching "start of line" and "end of
2255 line", PCRE treats the subject string as consisting of a single line of
2256 characters, even if it actually contains newlines. The "start of line"
2257 metacharacter (^) matches only at the start of the string, and the "end
2258 of line" metacharacter ($) matches only at the end of the string, or
2259 before a terminating newline (except when PCRE_DOLLAR_ENDONLY is set).
2260 Note, however, that unless PCRE_DOTALL is set, the "any character"
2261 metacharacter (.) does not match at a newline. This behaviour (for ^,
2262 $, and dot) is the same as Perl.
2264 When PCRE_MULTILINE it is set, the "start of line" and "end of line"
2265 constructs match immediately following or immediately before internal
2266 newlines in the subject string, respectively, as well as at the very
2267 start and end. This is equivalent to Perl's /m option, and it can be
2268 changed within a pattern by a (?m) option setting. If there are no new-
2269 lines in a subject string, or no occurrences of ^ or $ in a pattern,
2270 setting PCRE_MULTILINE has no effect.
2274 This option locks out interpretation of the pattern as UTF-8 (or UTF-16
2275 or UTF-32 in the 16-bit and 32-bit libraries). In particular, it pre-
2276 vents the creator of the pattern from switching to UTF interpretation
2277 by starting the pattern with (*UTF). This may be useful in applications
2278 that process patterns from external sources. The combination of
2279 PCRE_UTF8 and PCRE_NEVER_UTF also causes an error.
2284 PCRE_NEWLINE_ANYCRLF
2287 These options override the default newline definition that was chosen
2288 when PCRE was built. Setting the first or the second specifies that a
2289 newline is indicated by a single character (CR or LF, respectively).
2290 Setting PCRE_NEWLINE_CRLF specifies that a newline is indicated by the
2291 two-character CRLF sequence. Setting PCRE_NEWLINE_ANYCRLF specifies
2292 that any of the three preceding sequences should be recognized. Setting
2293 PCRE_NEWLINE_ANY specifies that any Unicode newline sequence should be
2296 In an ASCII/Unicode environment, the Unicode newline sequences are the
2297 three just mentioned, plus the single characters VT (vertical tab,
2298 U+000B), FF (form feed, U+000C), NEL (next line, U+0085), LS (line sep-
2299 arator, U+2028), and PS (paragraph separator, U+2029). For the 8-bit
2300 library, the last two are recognized only in UTF-8 mode.
2302 When PCRE is compiled to run in an EBCDIC (mainframe) environment, the
2303 code for CR is 0x0d, the same as ASCII. However, the character code for
2304 LF is normally 0x15, though in some EBCDIC environments 0x25 is used.
2305 Whichever of these is not LF is made to correspond to Unicode's NEL
2306 character. EBCDIC codes are all less than 256. For more details, see
2307 the pcrebuild documentation.
2309 The newline setting in the options word uses three bits that are
2310 treated as a number, giving eight possibilities. Currently only six are
2311 used (default plus the five values above). This means that if you set
2312 more than one newline option, the combination may or may not be sensi-
2313 ble. For example, PCRE_NEWLINE_CR with PCRE_NEWLINE_LF is equivalent to
2314 PCRE_NEWLINE_CRLF, but other combinations may yield unused numbers and
2317 The only time that a line break in a pattern is specially recognized
2318 when compiling is when PCRE_EXTENDED is set. CR and LF are white space
2319 characters, and so are ignored in this mode. Also, an unescaped # out-
2320 side a character class indicates a comment that lasts until after the
2321 next line break sequence. In other circumstances, line break sequences
2322 in patterns are treated as literal data.
2324 The newline option that is set at compile time becomes the default that
2325 is used for pcre_exec() and pcre_dfa_exec(), but it can be overridden.
2327 PCRE_NO_AUTO_CAPTURE
2329 If this option is set, it disables the use of numbered capturing paren-
2330 theses in the pattern. Any opening parenthesis that is not followed by
2331 ? behaves as if it were followed by ?: but named parentheses can still
2332 be used for capturing (and they acquire numbers in the usual way).
2333 There is no equivalent of this option in Perl.
2335 PCRE_NO_AUTO_POSSESS
2337 If this option is set, it disables "auto-possessification". This is an
2338 optimization that, for example, turns a+b into a++b in order to avoid
2339 backtracks into a+ that can never be successful. However, if callouts
2340 are in use, auto-possessification means that some of them are never
2341 taken. You can set this option if you want the matching functions to do
2342 a full unoptimized search and run all the callouts, but it is mainly
2343 provided for testing purposes.
2345 PCRE_NO_START_OPTIMIZE
2347 This is an option that acts at matching time; that is, it is really an
2348 option for pcre_exec() or pcre_dfa_exec(). If it is set at compile
2349 time, it is remembered with the compiled pattern and assumed at match-
2350 ing time. This is necessary if you want to use JIT execution, because
2351 the JIT compiler needs to know whether or not this option is set. For
2352 details see the discussion of PCRE_NO_START_OPTIMIZE below.
2356 This option changes the way PCRE processes \B, \b, \D, \d, \S, \s, \W,
2357 \w, and some of the POSIX character classes. By default, only ASCII
2358 characters are recognized, but if PCRE_UCP is set, Unicode properties
2359 are used instead to classify characters. More details are given in the
2360 section on generic character types in the pcrepattern page. If you set
2361 PCRE_UCP, matching one of the items it affects takes much longer. The
2362 option is available only if PCRE has been compiled with Unicode prop-
2367 This option inverts the "greediness" of the quantifiers so that they
2368 are not greedy by default, but become greedy if followed by "?". It is
2369 not compatible with Perl. It can also be set by a (?U) option setting
2374 This option causes PCRE to regard both the pattern and the subject as
2375 strings of UTF-8 characters instead of single-byte strings. However, it
2376 is available only when PCRE is built to include UTF support. If not,
2377 the use of this option provokes an error. Details of how this option
2378 changes the behaviour of PCRE are given in the pcreunicode page.
2382 When PCRE_UTF8 is set, the validity of the pattern as a UTF-8 string is
2383 automatically checked. There is a discussion about the validity of
2384 UTF-8 strings in the pcreunicode page. If an invalid UTF-8 sequence is
2385 found, pcre_compile() returns an error. If you already know that your
2386 pattern is valid, and you want to skip this check for performance rea-
2387 sons, you can set the PCRE_NO_UTF8_CHECK option. When it is set, the
2388 effect of passing an invalid UTF-8 string as a pattern is undefined. It
2389 may cause your program to crash or loop. Note that this option can also
2390 be passed to pcre_exec() and pcre_dfa_exec(), to suppress the validity
2391 checking of subject strings only. If the same string is being matched
2392 many times, the option can be safely set for the second and subsequent
2393 matchings to improve performance.
2396 COMPILATION ERROR CODES
2398 The following table lists the error codes than may be returned by
2399 pcre_compile2(), along with the error messages that may be returned by
2400 both compiling functions. Note that error messages are always 8-bit
2401 ASCII strings, even in 16-bit or 32-bit mode. As PCRE has developed,
2402 some error codes have fallen out of use. To avoid confusion, they have
2406 1 \ at end of pattern
2407 2 \c at end of pattern
2408 3 unrecognized character follows \
2409 4 numbers out of order in {} quantifier
2410 5 number too big in {} quantifier
2411 6 missing terminating ] for character class
2412 7 invalid escape sequence in character class
2413 8 range out of order in character class
2415 10 [this code is not in use]
2416 11 internal error: unexpected repeat
2417 12 unrecognized character after (? or (?-
2418 13 POSIX named classes are supported only within a class
2420 15 reference to non-existent subpattern
2421 16 erroffset passed as NULL
2422 17 unknown option bit(s) set
2423 18 missing ) after comment
2424 19 [this code is not in use]
2425 20 regular expression is too large
2426 21 failed to get memory
2427 22 unmatched parentheses
2428 23 internal error: code overflow
2429 24 unrecognized character after (?<
2430 25 lookbehind assertion is not fixed length
2431 26 malformed number or name after (?(
2432 27 conditional group contains more than two branches
2433 28 assertion expected after (?(
2434 29 (?R or (?[+-]digits must be followed by )
2435 30 unknown POSIX class name
2436 31 POSIX collating elements are not supported
2437 32 this version of PCRE is compiled without UTF support
2438 33 [this code is not in use]
2439 34 character value in \x{} or \o{} is too large
2440 35 invalid condition (?(0)
2441 36 \C not allowed in lookbehind assertion
2442 37 PCRE does not support \L, \l, \N{name}, \U, or \u
2443 38 number after (?C is > 255
2444 39 closing ) for (?C expected
2445 40 recursive call could loop indefinitely
2446 41 unrecognized character after (?P
2447 42 syntax error in subpattern name (missing terminator)
2448 43 two named subpatterns have the same name
2449 44 invalid UTF-8 string (specifically UTF-8)
2450 45 support for \P, \p, and \X has not been compiled
2451 46 malformed \P or \p sequence
2452 47 unknown property name after \P or \p
2453 48 subpattern name is too long (maximum 32 characters)
2454 49 too many named subpatterns (maximum 10000)
2455 50 [this code is not in use]
2456 51 octal value is greater than \377 in 8-bit non-UTF-8 mode
2457 52 internal error: overran compiling workspace
2458 53 internal error: previously-checked referenced subpattern
2460 54 DEFINE group contains more than one branch
2461 55 repeating a DEFINE group is not allowed
2462 56 inconsistent NEWLINE options
2463 57 \g is not followed by a braced, angle-bracketed, or quoted
2464 name/number or by a plain number
2465 58 a numbered reference must not be zero
2466 59 an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)
2467 60 (*VERB) not recognized or malformed
2468 61 number is too big
2469 62 subpattern name expected
2470 63 digit expected after (?+
2471 64 ] is an invalid data character in JavaScript compatibility mode
2472 65 different names for subpatterns of the same number are
2474 66 (*MARK) must have an argument
2475 67 this version of PCRE is not compiled with Unicode property
2477 68 \c must be followed by an ASCII character
2478 69 \k is not followed by a braced, angle-bracketed, or quoted name
2479 70 internal error: unknown opcode in find_fixedlength()
2480 71 \N is not supported in a class
2481 72 too many forward references
2482 73 disallowed Unicode code point (>= 0xd800 && <= 0xdfff)
2483 74 invalid UTF-16 string (specifically UTF-16)
2484 75 name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)
2485 76 character value in \u.... sequence is too large
2486 77 invalid UTF-32 string (specifically UTF-32)
2487 78 setting UTF is disabled by the application
2488 79 non-hex character in \x{} (closing brace missing?)
2489 80 non-octal character in \o{} (closing brace missing?)
2490 81 missing opening brace after \o
2491 82 parentheses are too deeply nested
2492 83 invalid range in character class
2493 84 group name must start with a non-digit
2494 85 parentheses are too deeply nested (stack check)
2496 The numbers 32 and 10000 in errors 48 and 49 are defaults; different
2497 values may be used if the limits were changed when PCRE was built.
2502 pcre_extra *pcre_study(const pcre *code, int options,
2503 const char **errptr);
2505 If a compiled pattern is going to be used several times, it is worth
2506 spending more time analyzing it in order to speed up the time taken for
2507 matching. The function pcre_study() takes a pointer to a compiled pat-
2508 tern as its first argument. If studying the pattern produces additional
2509 information that will help speed up matching, pcre_study() returns a
2510 pointer to a pcre_extra block, in which the study_data field points to
2511 the results of the study.
2513 The returned value from pcre_study() can be passed directly to
2514 pcre_exec() or pcre_dfa_exec(). However, a pcre_extra block also con-
2515 tains other fields that can be set by the caller before the block is
2516 passed; these are described below in the section on matching a pattern.
2518 If studying the pattern does not produce any useful information,
2519 pcre_study() returns NULL by default. In that circumstance, if the
2520 calling program wants to pass any of the other fields to pcre_exec() or
2521 pcre_dfa_exec(), it must set up its own pcre_extra block. However, if
2522 pcre_study() is called with the PCRE_STUDY_EXTRA_NEEDED option, it
2523 returns a pcre_extra block even if studying did not find any additional
2524 information. It may still return NULL, however, if an error occurs in
2527 The second argument of pcre_study() contains option bits. There are
2528 three further options in addition to PCRE_STUDY_EXTRA_NEEDED:
2530 PCRE_STUDY_JIT_COMPILE
2531 PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE
2532 PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE
2534 If any of these are set, and the just-in-time compiler is available,
2535 the pattern is further compiled into machine code that executes much
2536 faster than the pcre_exec() interpretive matching function. If the
2537 just-in-time compiler is not available, these options are ignored. All
2538 undefined bits in the options argument must be zero.
2540 JIT compilation is a heavyweight optimization. It can take some time
2541 for patterns to be analyzed, and for one-off matches and simple pat-
2542 terns the benefit of faster execution might be offset by a much slower
2543 study time. Not all patterns can be optimized by the JIT compiler. For
2544 those that cannot be handled, matching automatically falls back to the
2545 pcre_exec() interpreter. For more details, see the pcrejit documenta-
2548 The third argument for pcre_study() is a pointer for an error message.
2549 If studying succeeds (even if no data is returned), the variable it
2550 points to is set to NULL. Otherwise it is set to point to a textual
2551 error message. This is a static string that is part of the library. You
2552 must not try to free it. You should test the error pointer for NULL
2553 after calling pcre_study(), to be sure that it has run successfully.
2555 When you are finished with a pattern, you can free the memory used for
2556 the study data by calling pcre_free_study(). This function was added to
2557 the API for release 8.20. For earlier versions, the memory could be
2558 freed with pcre_free(), just like the pattern itself. This will still
2559 work in cases where JIT optimization is not used, but it is advisable
2560 to change to the new function when convenient.
2562 This is a typical way in which pcre_study() is used (except that in a
2563 real application there should be tests for errors):
2568 re = pcre_compile("pattern", 0, &error, &erroroffset, NULL);
2570 re, /* result of pcre_compile() */
2572 &error); /* set to NULL or points to a message */
2573 rc = pcre_exec( /* see below for details of pcre_exec() options */
2574 re, sd, "subject", 7, 0, 0, ovector, 30);
2576 pcre_free_study(sd);
2579 Studying a pattern does two things: first, a lower bound for the length
2580 of subject string that is needed to match the pattern is computed. This
2581 does not mean that there are any strings of that length that match, but
2582 it does guarantee that no shorter strings match. The value is used to
2583 avoid wasting time by trying to match strings that are shorter than the
2584 lower bound. You can find out the value in a calling program via the
2585 pcre_fullinfo() function.
2587 Studying a pattern is also useful for non-anchored patterns that do not
2588 have a single fixed starting character. A bitmap of possible starting
2589 bytes is created. This speeds up finding a position in the subject at
2590 which to start matching. (In 16-bit mode, the bitmap is used for 16-bit
2591 values less than 256. In 32-bit mode, the bitmap is used for 32-bit
2592 values less than 256.)
2594 These two optimizations apply to both pcre_exec() and pcre_dfa_exec(),
2595 and the information is also used by the JIT compiler. The optimiza-
2596 tions can be disabled by setting the PCRE_NO_START_OPTIMIZE option.
2597 You might want to do this if your pattern contains callouts or (*MARK)
2598 and you want to make use of these facilities in cases where matching
2601 PCRE_NO_START_OPTIMIZE can be specified at either compile time or exe-
2602 cution time. However, if PCRE_NO_START_OPTIMIZE is passed to
2603 pcre_exec(), (that is, after any JIT compilation has happened) JIT exe-
2604 cution is disabled. For JIT execution to work with PCRE_NO_START_OPTI-
2605 MIZE, the option must be set at compile time.
2607 There is a longer discussion of PCRE_NO_START_OPTIMIZE below.
2612 PCRE handles caseless matching, and determines whether characters are
2613 letters, digits, or whatever, by reference to a set of tables, indexed
2614 by character code point. When running in UTF-8 mode, or in the 16- or
2615 32-bit libraries, this applies only to characters with code points less
2616 than 256. By default, higher-valued code points never match escapes
2617 such as \w or \d. However, if PCRE is built with Unicode property sup-
2618 port, all characters can be tested with \p and \P, or, alternatively,
2619 the PCRE_UCP option can be set when a pattern is compiled; this causes
2620 \w and friends to use Unicode property support instead of the built-in
2623 The use of locales with Unicode is discouraged. If you are handling
2624 characters with code points greater than 128, you should either use
2625 Unicode support, or use locales, but not try to mix the two.
2627 PCRE contains an internal set of tables that are used when the final
2628 argument of pcre_compile() is NULL. These are sufficient for many
2629 applications. Normally, the internal tables recognize only ASCII char-
2630 acters. However, when PCRE is built, it is possible to cause the inter-
2631 nal tables to be rebuilt in the default "C" locale of the local system,
2632 which may cause them to be different.
2634 The internal tables can always be overridden by tables supplied by the
2635 application that calls PCRE. These may be created in a different locale
2636 from the default. As more and more applications change to using Uni-
2637 code, the need for this locale support is expected to die away.
2639 External tables are built by calling the pcre_maketables() function,
2640 which has no arguments, in the relevant locale. The result can then be
2641 passed to pcre_compile() as often as necessary. For example, to build
2642 and use tables that are appropriate for the French locale (where
2643 accented characters with values greater than 128 are treated as let-
2644 ters), the following code could be used:
2646 setlocale(LC_CTYPE, "fr_FR");
2647 tables = pcre_maketables();
2648 re = pcre_compile(..., tables);
2650 The locale name "fr_FR" is used on Linux and other Unix-like systems;
2651 if you are using Windows, the name for the French locale is "french".
2653 When pcre_maketables() runs, the tables are built in memory that is
2654 obtained via pcre_malloc. It is the caller's responsibility to ensure
2655 that the memory containing the tables remains available for as long as
2658 The pointer that is passed to pcre_compile() is saved with the compiled
2659 pattern, and the same tables are used via this pointer by pcre_study()
2660 and also by pcre_exec() and pcre_dfa_exec(). Thus, for any single pat-
2661 tern, compilation, studying and matching all happen in the same locale,
2662 but different patterns can be processed in different locales.
2664 It is possible to pass a table pointer or NULL (indicating the use of
2665 the internal tables) to pcre_exec() or pcre_dfa_exec() (see the discus-
2666 sion below in the section on matching a pattern). This facility is pro-
2667 vided for use with pre-compiled patterns that have been saved and
2668 reloaded. Character tables are not saved with patterns, so if a non-
2669 standard table was used at compile time, it must be provided again when
2670 the reloaded pattern is matched. Attempting to use this facility to
2671 match a pattern in a different locale from the one in which it was com-
2672 piled is likely to lead to anomalous (usually incorrect) results.
2675 INFORMATION ABOUT A PATTERN
2677 int pcre_fullinfo(const pcre *code, const pcre_extra *extra,
2678 int what, void *where);
2680 The pcre_fullinfo() function returns information about a compiled pat-
2681 tern. It replaces the pcre_info() function, which was removed from the
2682 library at version 8.30, after more than 10 years of obsolescence.
2684 The first argument for pcre_fullinfo() is a pointer to the compiled
2685 pattern. The second argument is the result of pcre_study(), or NULL if
2686 the pattern was not studied. The third argument specifies which piece
2687 of information is required, and the fourth argument is a pointer to a
2688 variable to receive the data. The yield of the function is zero for
2689 success, or one of the following negative numbers:
2691 PCRE_ERROR_NULL the argument code was NULL
2692 the argument where was NULL
2693 PCRE_ERROR_BADMAGIC the "magic number" was not found
2694 PCRE_ERROR_BADENDIANNESS the pattern was compiled with different
2696 PCRE_ERROR_BADOPTION the value of what was invalid
2697 PCRE_ERROR_UNSET the requested field is not set
2699 The "magic number" is placed at the start of each compiled pattern as
2700 an simple check against passing an arbitrary memory pointer. The endi-
2701 anness error can occur if a compiled pattern is saved and reloaded on a
2702 different host. Here is a typical call of pcre_fullinfo(), to obtain
2703 the length of the compiled pattern:
2708 re, /* result of pcre_compile() */
2709 sd, /* result of pcre_study(), or NULL */
2710 PCRE_INFO_SIZE, /* what is required */
2711 &length); /* where to put the data */
2713 The possible values for the third argument are defined in pcre.h, and
2716 PCRE_INFO_BACKREFMAX
2718 Return the number of the highest back reference in the pattern. The
2719 fourth argument should point to an int variable. Zero is returned if
2720 there are no back references.
2722 PCRE_INFO_CAPTURECOUNT
2724 Return the number of capturing subpatterns in the pattern. The fourth
2725 argument should point to an int variable.
2727 PCRE_INFO_DEFAULT_TABLES
2729 Return a pointer to the internal default character tables within PCRE.
2730 The fourth argument should point to an unsigned char * variable. This
2731 information call is provided for internal use by the pcre_study() func-
2732 tion. External callers can cause PCRE to use its internal tables by
2733 passing a NULL table pointer.
2735 PCRE_INFO_FIRSTBYTE (deprecated)
2737 Return information about the first data unit of any matched string, for
2738 a non-anchored pattern. The name of this option refers to the 8-bit
2739 library, where data units are bytes. The fourth argument should point
2740 to an int variable. Negative values are used for special cases. How-
2741 ever, this means that when the 32-bit library is in non-UTF-32 mode,
2742 the full 32-bit range of characters cannot be returned. For this rea-
2743 son, this value is deprecated; use PCRE_INFO_FIRSTCHARACTERFLAGS and
2744 PCRE_INFO_FIRSTCHARACTER instead.
2746 If there is a fixed first value, for example, the letter "c" from a
2747 pattern such as (cat|cow|coyote), its value is returned. In the 8-bit
2748 library, the value is always less than 256. In the 16-bit library the
2749 value can be up to 0xffff. In the 32-bit library the value can be up to
2752 If there is no fixed first value, and if either
2754 (a) the pattern was compiled with the PCRE_MULTILINE option, and every
2755 branch starts with "^", or
2757 (b) every branch of the pattern starts with ".*" and PCRE_DOTALL is not
2758 set (if it were set, the pattern would be anchored),
2760 -1 is returned, indicating that the pattern matches only at the start
2761 of a subject string or after any newline within the string. Otherwise
2762 -2 is returned. For anchored patterns, -2 is returned.
2764 PCRE_INFO_FIRSTCHARACTER
2766 Return the value of the first data unit (non-UTF character) of any
2767 matched string in the situation where PCRE_INFO_FIRSTCHARACTERFLAGS
2768 returns 1; otherwise return 0. The fourth argument should point to an
2771 In the 8-bit library, the value is always less than 256. In the 16-bit
2772 library the value can be up to 0xffff. In the 32-bit library in UTF-32
2773 mode the value can be up to 0x10ffff, and up to 0xffffffff when not
2776 PCRE_INFO_FIRSTCHARACTERFLAGS
2778 Return information about the first data unit of any matched string, for
2779 a non-anchored pattern. The fourth argument should point to an int
2782 If there is a fixed first value, for example, the letter "c" from a
2783 pattern such as (cat|cow|coyote), 1 is returned, and the character
2784 value can be retrieved using PCRE_INFO_FIRSTCHARACTER. If there is no
2785 fixed first value, and if either
2787 (a) the pattern was compiled with the PCRE_MULTILINE option, and every
2788 branch starts with "^", or
2790 (b) every branch of the pattern starts with ".*" and PCRE_DOTALL is not
2791 set (if it were set, the pattern would be anchored),
2793 2 is returned, indicating that the pattern matches only at the start of
2794 a subject string or after any newline within the string. Otherwise 0 is
2795 returned. For anchored patterns, 0 is returned.
2797 PCRE_INFO_FIRSTTABLE
2799 If the pattern was studied, and this resulted in the construction of a
2800 256-bit table indicating a fixed set of values for the first data unit
2801 in any matching string, a pointer to the table is returned. Otherwise
2802 NULL is returned. The fourth argument should point to an unsigned char
2807 Return 1 if the pattern contains any explicit matches for CR or LF
2808 characters, otherwise 0. The fourth argument should point to an int
2809 variable. An explicit match is either a literal CR or LF character, or
2814 Return 1 if the (?J) or (?-J) option setting is used in the pattern,
2815 otherwise 0. The fourth argument should point to an int variable. (?J)
2816 and (?-J) set and unset the local PCRE_DUPNAMES option, respectively.
2820 Return 1 if the pattern was studied with one of the JIT options, and
2821 just-in-time compiling was successful. The fourth argument should point
2822 to an int variable. A return value of 0 means that JIT support is not
2823 available in this version of PCRE, or that the pattern was not studied
2824 with a JIT option, or that the JIT compiler could not handle this par-
2825 ticular pattern. See the pcrejit documentation for details of what can
2826 and cannot be handled.
2830 If the pattern was successfully studied with a JIT option, return the
2831 size of the JIT compiled code, otherwise return zero. The fourth argu-
2832 ment should point to a size_t variable.
2834 PCRE_INFO_LASTLITERAL
2836 Return the value of the rightmost literal data unit that must exist in
2837 any matched string, other than at its start, if such a value has been
2838 recorded. The fourth argument should point to an int variable. If there
2839 is no such value, -1 is returned. For anchored patterns, a last literal
2840 value is recorded only if it follows something of variable length. For
2841 example, for the pattern /^a\d+z\d+/ the returned value is "z", but for
2842 /^a\dz\d/ the returned value is -1.
2844 Since for the 32-bit library using the non-UTF-32 mode, this function
2845 is unable to return the full 32-bit range of characters, this value is
2846 deprecated; instead the PCRE_INFO_REQUIREDCHARFLAGS and
2847 PCRE_INFO_REQUIREDCHAR values should be used.
2849 PCRE_INFO_MATCH_EMPTY
2851 Return 1 if the pattern can match an empty string, otherwise 0. The
2852 fourth argument should point to an int variable.
2854 PCRE_INFO_MATCHLIMIT
2856 If the pattern set a match limit by including an item of the form
2857 (*LIMIT_MATCH=nnnn) at the start, the value is returned. The fourth
2858 argument should point to an unsigned 32-bit integer. If no such value
2859 has been set, the call to pcre_fullinfo() returns the error
2862 PCRE_INFO_MAXLOOKBEHIND
2864 Return the number of characters (NB not data units) in the longest
2865 lookbehind assertion in the pattern. This information is useful when
2866 doing multi-segment matching using the partial matching facilities.
2867 Note that the simple assertions \b and \B require a one-character look-
2868 behind. \A also registers a one-character lookbehind, though it does
2869 not actually inspect the previous character. This is to ensure that at
2870 least one character from the old segment is retained when a new segment
2871 is processed. Otherwise, if there are no lookbehinds in the pattern, \A
2872 might match incorrectly at the start of a new segment.
2876 If the pattern was studied and a minimum length for matching subject
2877 strings was computed, its value is returned. Otherwise the returned
2878 value is -1. The value is a number of characters, which in UTF mode may
2879 be different from the number of data units. The fourth argument should
2880 point to an int variable. A non-negative value is a lower bound to the
2881 length of any matching string. There may not be any strings of that
2882 length that do actually match, but every string that does match is at
2886 PCRE_INFO_NAMEENTRYSIZE
2889 PCRE supports the use of named as well as numbered capturing parenthe-
2890 ses. The names are just an additional way of identifying the parenthe-
2891 ses, which still acquire numbers. Several convenience functions such as
2892 pcre_get_named_substring() are provided for extracting captured sub-
2893 strings by name. It is also possible to extract the data directly, by
2894 first converting the name to a number in order to access the correct
2895 pointers in the output vector (described with pcre_exec() below). To do
2896 the conversion, you need to use the name-to-number map, which is
2897 described by these three values.
2899 The map consists of a number of fixed-size entries. PCRE_INFO_NAMECOUNT
2900 gives the number of entries, and PCRE_INFO_NAMEENTRYSIZE gives the size
2901 of each entry; both of these return an int value. The entry size
2902 depends on the length of the longest name. PCRE_INFO_NAMETABLE returns
2903 a pointer to the first entry of the table. This is a pointer to char in
2904 the 8-bit library, where the first two bytes of each entry are the num-
2905 ber of the capturing parenthesis, most significant byte first. In the
2906 16-bit library, the pointer points to 16-bit data units, the first of
2907 which contains the parenthesis number. In the 32-bit library, the
2908 pointer points to 32-bit data units, the first of which contains the
2909 parenthesis number. The rest of the entry is the corresponding name,
2912 The names are in alphabetical order. If (?| is used to create multiple
2913 groups with the same number, as described in the section on duplicate
2914 subpattern numbers in the pcrepattern page, the groups may be given the
2915 same name, but there is only one entry in the table. Different names
2916 for groups of the same number are not permitted. Duplicate names for
2917 subpatterns with different numbers are permitted, but only if PCRE_DUP-
2918 NAMES is set. They appear in the table in the order in which they were
2919 found in the pattern. In the absence of (?| this is the order of
2920 increasing number; when (?| is used this is not necessarily the case
2921 because later subpatterns may have lower numbers.
2923 As a simple example of the name/number table, consider the following
2924 pattern after compilation by the 8-bit library (assume PCRE_EXTENDED is
2925 set, so white space - including newlines - is ignored):
2927 (?<date> (?<year>(\d\d)?\d\d) -
2928 (?<month>\d\d) - (?<day>\d\d) )
2930 There are four named subpatterns, so the table has four entries, and
2931 each entry in the table is eight bytes long. The table is as follows,
2932 with non-printing bytes shows in hexadecimal, and undefined bytes shown
2936 00 05 d a y 00 ?? ??
2940 When writing code to extract data from named subpatterns using the
2941 name-to-number map, remember that the length of the entries is likely
2942 to be different for each compiled pattern.
2946 Return 1 if the pattern can be used for partial matching with
2947 pcre_exec(), otherwise 0. The fourth argument should point to an int
2948 variable. From release 8.00, this always returns 1, because the
2949 restrictions that previously applied to partial matching have been
2950 lifted. The pcrepartial documentation gives details of partial match-
2955 Return a copy of the options with which the pattern was compiled. The
2956 fourth argument should point to an unsigned long int variable. These
2957 option bits are those specified in the call to pcre_compile(), modified
2958 by any top-level option settings at the start of the pattern itself. In
2959 other words, they are the options that will be in force when matching
2960 starts. For example, if the pattern /(?im)abc(?-i)d/ is compiled with
2961 the PCRE_EXTENDED option, the result is PCRE_CASELESS, PCRE_MULTILINE,
2964 A pattern is automatically anchored by PCRE if all of its top-level
2965 alternatives begin with one of the following:
2967 ^ unless PCRE_MULTILINE is set
2970 .* if PCRE_DOTALL is set and there are no back
2971 references to the subpattern in which .* appears
2973 For such patterns, the PCRE_ANCHORED bit is set in the options returned
2976 PCRE_INFO_RECURSIONLIMIT
2978 If the pattern set a recursion limit by including an item of the form
2979 (*LIMIT_RECURSION=nnnn) at the start, the value is returned. The fourth
2980 argument should point to an unsigned 32-bit integer. If no such value
2981 has been set, the call to pcre_fullinfo() returns the error
2986 Return the size of the compiled pattern in bytes (for all three
2987 libraries). The fourth argument should point to a size_t variable. This
2988 value does not include the size of the pcre structure that is returned
2989 by pcre_compile(). The value that is passed as the argument to
2990 pcre_malloc() when pcre_compile() is getting memory in which to place
2991 the compiled data is the value returned by this option plus the size of
2992 the pcre structure. Studying a compiled pattern, with or without JIT,
2993 does not alter the value returned by this option.
2997 Return the size in bytes (for all three libraries) of the data block
2998 pointed to by the study_data field in a pcre_extra block. If pcre_extra
2999 is NULL, or there is no study data, zero is returned. The fourth argu-
3000 ment should point to a size_t variable. The study_data field is set by
3001 pcre_study() to record information that will speed up matching (see the
3002 section entitled "Studying a pattern" above). The format of the
3003 study_data block is private, but its length is made available via this
3004 option so that it can be saved and restored (see the pcreprecompile
3005 documentation for details).
3007 PCRE_INFO_REQUIREDCHARFLAGS
3009 Returns 1 if there is a rightmost literal data unit that must exist in
3010 any matched string, other than at its start. The fourth argument should
3011 point to an int variable. If there is no such value, 0 is returned. If
3012 returning 1, the character value itself can be retrieved using
3013 PCRE_INFO_REQUIREDCHAR.
3015 For anchored patterns, a last literal value is recorded only if it fol-
3016 lows something of variable length. For example, for the pattern
3017 /^a\d+z\d+/ the returned value 1 (with "z" returned from
3018 PCRE_INFO_REQUIREDCHAR), but for /^a\dz\d/ the returned value is 0.
3020 PCRE_INFO_REQUIREDCHAR
3022 Return the value of the rightmost literal data unit that must exist in
3023 any matched string, other than at its start, if such a value has been
3024 recorded. The fourth argument should point to an uint32_t variable. If
3025 there is no such value, 0 is returned.
3030 int pcre_refcount(pcre *code, int adjust);
3032 The pcre_refcount() function is used to maintain a reference count in
3033 the data block that contains a compiled pattern. It is provided for the
3034 benefit of applications that operate in an object-oriented manner,
3035 where different parts of the application may be using the same compiled
3036 pattern, but you want to free the block when they are all done.
3038 When a pattern is compiled, the reference count field is initialized to
3039 zero. It is changed only by calling this function, whose action is to
3040 add the adjust value (which may be positive or negative) to it. The
3041 yield of the function is the new value. However, the value of the count
3042 is constrained to lie between 0 and 65535, inclusive. If the new value
3043 is outside these limits, it is forced to the appropriate limit value.
3045 Except when it is zero, the reference count is not correctly preserved
3046 if a pattern is compiled on one host and then transferred to a host
3047 whose byte-order is different. (This seems a highly unlikely scenario.)
3050 MATCHING A PATTERN: THE TRADITIONAL FUNCTION
3052 int pcre_exec(const pcre *code, const pcre_extra *extra,
3053 const char *subject, int length, int startoffset,
3054 int options, int *ovector, int ovecsize);
3056 The function pcre_exec() is called to match a subject string against a
3057 compiled pattern, which is passed in the code argument. If the pattern
3058 was studied, the result of the study should be passed in the extra
3059 argument. You can call pcre_exec() with the same code and extra argu-
3060 ments as many times as you like, in order to match different subject
3061 strings with the same pattern.
3063 This function is the main matching facility of the library, and it
3064 operates in a Perl-like manner. For specialist use there is also an
3065 alternative matching function, which is described below in the section
3066 about the pcre_dfa_exec() function.
3068 In most applications, the pattern will have been compiled (and option-
3069 ally studied) in the same process that calls pcre_exec(). However, it
3070 is possible to save compiled patterns and study data, and then use them
3071 later in different processes, possibly even on different hosts. For a
3072 discussion about this, see the pcreprecompile documentation.
3074 Here is an example of a simple call to pcre_exec():
3079 re, /* result of pcre_compile() */
3080 NULL, /* we didn't study the pattern */
3081 "some string", /* the subject string */
3082 11, /* the length of the subject string */
3083 0, /* start at offset 0 in the subject */
3084 0, /* default options */
3085 ovector, /* vector of integers for substring information */
3086 30); /* number of elements (NOT size in bytes) */
3088 Extra data for pcre_exec()
3090 If the extra argument is not NULL, it must point to a pcre_extra data
3091 block. The pcre_study() function returns such a block (when it doesn't
3092 return NULL), but you can also create one for yourself, and pass addi-
3093 tional information in it. The pcre_extra block contains the following
3094 fields (not necessarily in this order):
3096 unsigned long int flags;
3098 void *executable_jit;
3099 unsigned long int match_limit;
3100 unsigned long int match_limit_recursion;
3102 const unsigned char *tables;
3103 unsigned char **mark;
3105 In the 16-bit version of this structure, the mark field has type
3108 In the 32-bit version of this structure, the mark field has type
3111 The flags field is used to specify which of the other fields are set.
3114 PCRE_EXTRA_CALLOUT_DATA
3115 PCRE_EXTRA_EXECUTABLE_JIT
3117 PCRE_EXTRA_MATCH_LIMIT
3118 PCRE_EXTRA_MATCH_LIMIT_RECURSION
3119 PCRE_EXTRA_STUDY_DATA
3122 Other flag bits should be set to zero. The study_data field and some-
3123 times the executable_jit field are set in the pcre_extra block that is
3124 returned by pcre_study(), together with the appropriate flag bits. You
3125 should not set these yourself, but you may add to the block by setting
3126 other fields and their corresponding flag bits.
3128 The match_limit field provides a means of preventing PCRE from using up
3129 a vast amount of resources when running patterns that are not going to
3130 match, but which have a very large number of possibilities in their
3131 search trees. The classic example is a pattern that uses nested unlim-
3134 Internally, pcre_exec() uses a function called match(), which it calls
3135 repeatedly (sometimes recursively). The limit set by match_limit is
3136 imposed on the number of times this function is called during a match,
3137 which has the effect of limiting the amount of backtracking that can
3138 take place. For patterns that are not anchored, the count restarts from
3139 zero for each position in the subject string.
3141 When pcre_exec() is called with a pattern that was successfully studied
3142 with a JIT option, the way that the matching is executed is entirely
3143 different. However, there is still the possibility of runaway matching
3144 that goes on for a very long time, and so the match_limit value is also
3145 used in this case (but in a different way) to limit how long the match-
3148 The default value for the limit can be set when PCRE is built; the
3149 default default is 10 million, which handles all but the most extreme
3150 cases. You can override the default by suppling pcre_exec() with a
3151 pcre_extra block in which match_limit is set, and
3152 PCRE_EXTRA_MATCH_LIMIT is set in the flags field. If the limit is
3153 exceeded, pcre_exec() returns PCRE_ERROR_MATCHLIMIT.
3155 A value for the match limit may also be supplied by an item at the
3156 start of a pattern of the form
3160 where d is a decimal number. However, such a setting is ignored unless
3161 d is less than the limit set by the caller of pcre_exec() or, if no
3162 such limit is set, less than the default.
3164 The match_limit_recursion field is similar to match_limit, but instead
3165 of limiting the total number of times that match() is called, it limits
3166 the depth of recursion. The recursion depth is a smaller number than
3167 the total number of calls, because not all calls to match() are recur-
3168 sive. This limit is of use only if it is set smaller than match_limit.
3170 Limiting the recursion depth limits the amount of machine stack that
3171 can be used, or, when PCRE has been compiled to use memory on the heap
3172 instead of the stack, the amount of heap memory that can be used. This
3173 limit is not relevant, and is ignored, when matching is done using JIT
3176 The default value for match_limit_recursion can be set when PCRE is
3177 built; the default default is the same value as the default for
3178 match_limit. You can override the default by suppling pcre_exec() with
3179 a pcre_extra block in which match_limit_recursion is set, and
3180 PCRE_EXTRA_MATCH_LIMIT_RECURSION is set in the flags field. If the
3181 limit is exceeded, pcre_exec() returns PCRE_ERROR_RECURSIONLIMIT.
3183 A value for the recursion limit may also be supplied by an item at the
3184 start of a pattern of the form
3186 (*LIMIT_RECURSION=d)
3188 where d is a decimal number. However, such a setting is ignored unless
3189 d is less than the limit set by the caller of pcre_exec() or, if no
3190 such limit is set, less than the default.
3192 The callout_data field is used in conjunction with the "callout" fea-
3193 ture, and is described in the pcrecallout documentation.
3195 The tables field is provided for use with patterns that have been pre-
3196 compiled using custom character tables, saved to disc or elsewhere, and
3197 then reloaded, because the tables that were used to compile a pattern
3198 are not saved with it. See the pcreprecompile documentation for a dis-
3199 cussion of saving compiled patterns for later use. If NULL is passed
3200 using this mechanism, it forces PCRE's internal tables to be used.
3202 Warning: The tables that pcre_exec() uses must be the same as those
3203 that were used when the pattern was compiled. If this is not the case,
3204 the behaviour of pcre_exec() is undefined. Therefore, when a pattern is
3205 compiled and matched in the same process, this field should never be
3206 set. In this (the most common) case, the correct table pointer is auto-
3207 matically passed with the compiled pattern from pcre_compile() to
3210 If PCRE_EXTRA_MARK is set in the flags field, the mark field must be
3211 set to point to a suitable variable. If the pattern contains any back-
3212 tracking control verbs such as (*MARK:NAME), and the execution ends up
3213 with a name to pass back, a pointer to the name string (zero termi-
3214 nated) is placed in the variable pointed to by the mark field. The
3215 names are within the compiled pattern; if you wish to retain such a
3216 name you must copy it before freeing the memory of a compiled pattern.
3217 If there is no name to pass back, the variable pointed to by the mark
3218 field is set to NULL. For details of the backtracking control verbs,
3219 see the section entitled "Backtracking control" in the pcrepattern doc-
3222 Option bits for pcre_exec()
3224 The unused bits of the options argument for pcre_exec() must be zero.
3225 The only bits that may be set are PCRE_ANCHORED, PCRE_NEWLINE_xxx,
3226 PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART,
3227 PCRE_NO_START_OPTIMIZE, PCRE_NO_UTF8_CHECK, PCRE_PARTIAL_HARD, and
3230 If the pattern was successfully studied with one of the just-in-time
3231 (JIT) compile options, the only supported options for JIT execution are
3232 PCRE_NO_UTF8_CHECK, PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY,
3233 PCRE_NOTEMPTY_ATSTART, PCRE_PARTIAL_HARD, and PCRE_PARTIAL_SOFT. If an
3234 unsupported option is used, JIT execution is disabled and the normal
3235 interpretive code in pcre_exec() is run.
3239 The PCRE_ANCHORED option limits pcre_exec() to matching at the first
3240 matching position. If a pattern was compiled with PCRE_ANCHORED, or
3241 turned out to be anchored by virtue of its contents, it cannot be made
3242 unachored at matching time.
3247 These options (which are mutually exclusive) control what the \R escape
3248 sequence matches. The choice is either to match only CR, LF, or CRLF,
3249 or to match any Unicode newline sequence. These options override the
3250 choice that was made or defaulted when the pattern was compiled.
3255 PCRE_NEWLINE_ANYCRLF
3258 These options override the newline definition that was chosen or
3259 defaulted when the pattern was compiled. For details, see the descrip-
3260 tion of pcre_compile() above. During matching, the newline choice
3261 affects the behaviour of the dot, circumflex, and dollar metacharac-
3262 ters. It may also alter the way the match position is advanced after a
3263 match failure for an unanchored pattern.
3265 When PCRE_NEWLINE_CRLF, PCRE_NEWLINE_ANYCRLF, or PCRE_NEWLINE_ANY is
3266 set, and a match attempt for an unanchored pattern fails when the cur-
3267 rent position is at a CRLF sequence, and the pattern contains no
3268 explicit matches for CR or LF characters, the match position is
3269 advanced by two characters instead of one, in other words, to after the
3272 The above rule is a compromise that makes the most common cases work as
3273 expected. For example, if the pattern is .+A (and the PCRE_DOTALL
3274 option is not set), it does not match the string "\r\nA" because, after
3275 failing at the start, it skips both the CR and the LF before retrying.
3276 However, the pattern [\r\n]A does match that string, because it con-
3277 tains an explicit CR or LF reference, and so advances only by one char-
3278 acter after the first failure.
3280 An explicit match for CR of LF is either a literal appearance of one of
3281 those characters, or one of the \r or \n escape sequences. Implicit
3282 matches such as [^X] do not count, nor does \s (which includes CR and
3283 LF in the characters that it matches).
3285 Notwithstanding the above, anomalous effects may still occur when CRLF
3286 is a valid newline sequence and explicit \r or \n escapes appear in the
3291 This option specifies that first character of the subject string is not
3292 the beginning of a line, so the circumflex metacharacter should not
3293 match before it. Setting this without PCRE_MULTILINE (at compile time)
3294 causes circumflex never to match. This option affects only the behav-
3295 iour of the circumflex metacharacter. It does not affect \A.
3299 This option specifies that the end of the subject string is not the end
3300 of a line, so the dollar metacharacter should not match it nor (except
3301 in multiline mode) a newline immediately before it. Setting this with-
3302 out PCRE_MULTILINE (at compile time) causes dollar never to match. This
3303 option affects only the behaviour of the dollar metacharacter. It does
3304 not affect \Z or \z.
3308 An empty string is not considered to be a valid match if this option is
3309 set. If there are alternatives in the pattern, they are tried. If all
3310 the alternatives match the empty string, the entire match fails. For
3311 example, if the pattern
3315 is applied to a string not beginning with "a" or "b", it matches an
3316 empty string at the start of the subject. With PCRE_NOTEMPTY set, this
3317 match is not valid, so PCRE searches further into the string for occur-
3318 rences of "a" or "b".
3320 PCRE_NOTEMPTY_ATSTART
3322 This is like PCRE_NOTEMPTY, except that an empty string match that is
3323 not at the start of the subject is permitted. If the pattern is
3324 anchored, such a match can occur only if the pattern contains \K.
3326 Perl has no direct equivalent of PCRE_NOTEMPTY or
3327 PCRE_NOTEMPTY_ATSTART, but it does make a special case of a pattern
3328 match of the empty string within its split() function, and when using
3329 the /g modifier. It is possible to emulate Perl's behaviour after
3330 matching a null string by first trying the match again at the same off-
3331 set with PCRE_NOTEMPTY_ATSTART and PCRE_ANCHORED, and then if that
3332 fails, by advancing the starting offset (see below) and trying an ordi-
3333 nary match again. There is some code that demonstrates how to do this
3334 in the pcredemo sample program. In the most general case, you have to
3335 check to see if the newline convention recognizes CRLF as a newline,
3336 and if so, and the current character is CR followed by LF, advance the
3337 starting offset by two characters instead of one.
3339 PCRE_NO_START_OPTIMIZE
3341 There are a number of optimizations that pcre_exec() uses at the start
3342 of a match, in order to speed up the process. For example, if it is
3343 known that an unanchored match must start with a specific character, it
3344 searches the subject for that character, and fails immediately if it
3345 cannot find it, without actually running the main matching function.
3346 This means that a special item such as (*COMMIT) at the start of a pat-
3347 tern is not considered until after a suitable starting point for the
3348 match has been found. Also, when callouts or (*MARK) items are in use,
3349 these "start-up" optimizations can cause them to be skipped if the pat-
3350 tern is never actually used. The start-up optimizations are in effect a
3351 pre-scan of the subject that takes place before the pattern is run.
3353 The PCRE_NO_START_OPTIMIZE option disables the start-up optimizations,
3354 possibly causing performance to suffer, but ensuring that in cases
3355 where the result is "no match", the callouts do occur, and that items
3356 such as (*COMMIT) and (*MARK) are considered at every possible starting
3357 position in the subject string. If PCRE_NO_START_OPTIMIZE is set at
3358 compile time, it cannot be unset at matching time. The use of
3359 PCRE_NO_START_OPTIMIZE at matching time (that is, passing it to
3360 pcre_exec()) disables JIT execution; in this situation, matching is
3361 always done using interpretively.
3363 Setting PCRE_NO_START_OPTIMIZE can change the outcome of a matching
3364 operation. Consider the pattern
3368 When this is compiled, PCRE records the fact that a match must start
3369 with the character "A". Suppose the subject string is "DEFABC". The
3370 start-up optimization scans along the subject, finds "A" and runs the
3371 first match attempt from there. The (*COMMIT) item means that the pat-
3372 tern must match the current starting position, which in this case, it
3373 does. However, if the same match is run with PCRE_NO_START_OPTIMIZE
3374 set, the initial scan along the subject string does not happen. The
3375 first match attempt is run starting from "D" and when this fails,
3376 (*COMMIT) prevents any further matches being tried, so the overall
3377 result is "no match". If the pattern is studied, more start-up opti-
3378 mizations may be used. For example, a minimum length for the subject
3379 may be recorded. Consider the pattern
3383 The minimum length for a match is one character. If the subject is
3384 "ABC", there will be attempts to match "ABC", "BC", "C", and then
3385 finally an empty string. If the pattern is studied, the final attempt
3386 does not take place, because PCRE knows that the subject is too short,
3387 and so the (*MARK) is never encountered. In this case, studying the
3388 pattern does not affect the overall match result, which is still "no
3389 match", but it does affect the auxiliary information that is returned.
3393 When PCRE_UTF8 is set at compile time, the validity of the subject as a
3394 UTF-8 string is automatically checked when pcre_exec() is subsequently
3395 called. The entire string is checked before any other processing takes
3396 place. The value of startoffset is also checked to ensure that it
3397 points to the start of a UTF-8 character. There is a discussion about
3398 the validity of UTF-8 strings in the pcreunicode page. If an invalid
3399 sequence of bytes is found, pcre_exec() returns the error
3400 PCRE_ERROR_BADUTF8 or, if PCRE_PARTIAL_HARD is set and the problem is a
3401 truncated character at the end of the subject, PCRE_ERROR_SHORTUTF8. In
3402 both cases, information about the precise nature of the error may also
3403 be returned (see the descriptions of these errors in the section enti-
3404 tled Error return values from pcre_exec() below). If startoffset con-
3405 tains a value that does not point to the start of a UTF-8 character (or
3406 to the end of the subject), PCRE_ERROR_BADUTF8_OFFSET is returned.
3408 If you already know that your subject is valid, and you want to skip
3409 these checks for performance reasons, you can set the
3410 PCRE_NO_UTF8_CHECK option when calling pcre_exec(). You might want to
3411 do this for the second and subsequent calls to pcre_exec() if you are
3412 making repeated calls to find all the matches in a single subject
3413 string. However, you should be sure that the value of startoffset
3414 points to the start of a character (or the end of the subject). When
3415 PCRE_NO_UTF8_CHECK is set, the effect of passing an invalid string as a
3416 subject or an invalid value of startoffset is undefined. Your program
3422 These options turn on the partial matching feature. For backwards com-
3423 patibility, PCRE_PARTIAL is a synonym for PCRE_PARTIAL_SOFT. A partial
3424 match occurs if the end of the subject string is reached successfully,
3425 but there are not enough subject characters to complete the match. If
3426 this happens when PCRE_PARTIAL_SOFT (but not PCRE_PARTIAL_HARD) is set,
3427 matching continues by testing any remaining alternatives. Only if no
3428 complete match can be found is PCRE_ERROR_PARTIAL returned instead of
3429 PCRE_ERROR_NOMATCH. In other words, PCRE_PARTIAL_SOFT says that the
3430 caller is prepared to handle a partial match, but only if no complete
3433 If PCRE_PARTIAL_HARD is set, it overrides PCRE_PARTIAL_SOFT. In this
3434 case, if a partial match is found, pcre_exec() immediately returns
3435 PCRE_ERROR_PARTIAL, without considering any other alternatives. In
3436 other words, when PCRE_PARTIAL_HARD is set, a partial match is consid-
3437 ered to be more important that an alternative complete match.
3439 In both cases, the portion of the string that was inspected when the
3440 partial match was found is set as the first matching string. There is a
3441 more detailed discussion of partial and multi-segment matching, with
3442 examples, in the pcrepartial documentation.
3444 The string to be matched by pcre_exec()
3446 The subject string is passed to pcre_exec() as a pointer in subject, a
3447 length in length, and a starting offset in startoffset. The units for
3448 length and startoffset are bytes for the 8-bit library, 16-bit data
3449 items for the 16-bit library, and 32-bit data items for the 32-bit
3452 If startoffset is negative or greater than the length of the subject,
3453 pcre_exec() returns PCRE_ERROR_BADOFFSET. When the starting offset is
3454 zero, the search for a match starts at the beginning of the subject,
3455 and this is by far the most common case. In UTF-8 or UTF-16 mode, the
3456 offset must point to the start of a character, or the end of the sub-
3457 ject (in UTF-32 mode, one data unit equals one character, so all off-
3458 sets are valid). Unlike the pattern string, the subject may contain
3461 A non-zero starting offset is useful when searching for another match
3462 in the same subject by calling pcre_exec() again after a previous suc-
3463 cess. Setting startoffset differs from just passing over a shortened
3464 string and setting PCRE_NOTBOL in the case of a pattern that begins
3465 with any kind of lookbehind. For example, consider the pattern
3469 which finds occurrences of "iss" in the middle of words. (\B matches
3470 only if the current position in the subject is not a word boundary.)
3471 When applied to the string "Mississipi" the first call to pcre_exec()
3472 finds the first occurrence. If pcre_exec() is called again with just
3473 the remainder of the subject, namely "issipi", it does not match,
3474 because \B is always false at the start of the subject, which is deemed
3475 to be a word boundary. However, if pcre_exec() is passed the entire
3476 string again, but with startoffset set to 4, it finds the second occur-
3477 rence of "iss" because it is able to look behind the starting point to
3478 discover that it is preceded by a letter.
3480 Finding all the matches in a subject is tricky when the pattern can
3481 match an empty string. It is possible to emulate Perl's /g behaviour by
3482 first trying the match again at the same offset, with the
3483 PCRE_NOTEMPTY_ATSTART and PCRE_ANCHORED options, and then if that
3484 fails, advancing the starting offset and trying an ordinary match
3485 again. There is some code that demonstrates how to do this in the pcre-
3486 demo sample program. In the most general case, you have to check to see
3487 if the newline convention recognizes CRLF as a newline, and if so, and
3488 the current character is CR followed by LF, advance the starting offset
3489 by two characters instead of one.
3491 If a non-zero starting offset is passed when the pattern is anchored,
3492 one attempt to match at the given offset is made. This can only succeed
3493 if the pattern does not require the match to be at the start of the
3496 How pcre_exec() returns captured substrings
3498 In general, a pattern matches a certain portion of the subject, and in
3499 addition, further substrings from the subject may be picked out by
3500 parts of the pattern. Following the usage in Jeffrey Friedl's book,
3501 this is called "capturing" in what follows, and the phrase "capturing
3502 subpattern" is used for a fragment of a pattern that picks out a sub-
3503 string. PCRE supports several other kinds of parenthesized subpattern
3504 that do not cause substrings to be captured.
3506 Captured substrings are returned to the caller via a vector of integers
3507 whose address is passed in ovector. The number of elements in the vec-
3508 tor is passed in ovecsize, which must be a non-negative number. Note:
3509 this argument is NOT the size of ovector in bytes.
3511 The first two-thirds of the vector is used to pass back captured sub-
3512 strings, each substring using a pair of integers. The remaining third
3513 of the vector is used as workspace by pcre_exec() while matching cap-
3514 turing subpatterns, and is not available for passing back information.
3515 The number passed in ovecsize should always be a multiple of three. If
3516 it is not, it is rounded down.
3518 When a match is successful, information about captured substrings is
3519 returned in pairs of integers, starting at the beginning of ovector,
3520 and continuing up to two-thirds of its length at the most. The first
3521 element of each pair is set to the offset of the first character in a
3522 substring, and the second is set to the offset of the first character
3523 after the end of a substring. These values are always data unit off-
3524 sets, even in UTF mode. They are byte offsets in the 8-bit library,
3525 16-bit data item offsets in the 16-bit library, and 32-bit data item
3526 offsets in the 32-bit library. Note: they are not character counts.
3528 The first pair of integers, ovector[0] and ovector[1], identify the
3529 portion of the subject string matched by the entire pattern. The next
3530 pair is used for the first capturing subpattern, and so on. The value
3531 returned by pcre_exec() is one more than the highest numbered pair that
3532 has been set. For example, if two substrings have been captured, the
3533 returned value is 3. If there are no capturing subpatterns, the return
3534 value from a successful match is 1, indicating that just the first pair
3535 of offsets has been set.
3537 If a capturing subpattern is matched repeatedly, it is the last portion
3538 of the string that it matched that is returned.
3540 If the vector is too small to hold all the captured substring offsets,
3541 it is used as far as possible (up to two-thirds of its length), and the
3542 function returns a value of zero. If neither the actual string matched
3543 nor any captured substrings are of interest, pcre_exec() may be called
3544 with ovector passed as NULL and ovecsize as zero. However, if the pat-
3545 tern contains back references and the ovector is not big enough to
3546 remember the related substrings, PCRE has to get additional memory for
3547 use during matching. Thus it is usually advisable to supply an ovector
3550 There are some cases where zero is returned (indicating vector over-
3551 flow) when in fact the vector is exactly the right size for the final
3552 match. For example, consider the pattern
3556 If a vector of 6 elements (allowing for only 1 captured substring) is
3557 given with subject string "abd", pcre_exec() will try to set the second
3558 captured string, thereby recording a vector overflow, before failing to
3559 match "c" and backing up to try the second alternative. The zero
3560 return, however, does correctly indicate that the maximum number of
3561 slots (namely 2) have been filled. In similar cases where there is tem-
3562 porary overflow, but the final number of used slots is actually less
3563 than the maximum, a non-zero value is returned.
3565 The pcre_fullinfo() function can be used to find out how many capturing
3566 subpatterns there are in a compiled pattern. The smallest size for
3567 ovector that will allow for n captured substrings, in addition to the
3568 offsets of the substring matched by the whole pattern, is (n+1)*3.
3570 It is possible for capturing subpattern number n+1 to match some part
3571 of the subject when subpattern n has not been used at all. For example,
3572 if the string "abc" is matched against the pattern (a|(z))(bc) the
3573 return from the function is 4, and subpatterns 1 and 3 are matched, but
3574 2 is not. When this happens, both values in the offset pairs corre-
3575 sponding to unused subpatterns are set to -1.
3577 Offset values that correspond to unused subpatterns at the end of the
3578 expression are also set to -1. For example, if the string "abc" is
3579 matched against the pattern (abc)(x(yz)?)? subpatterns 2 and 3 are not
3580 matched. The return from the function is 2, because the highest used
3581 capturing subpattern number is 1, and the offsets for for the second
3582 and third capturing subpatterns (assuming the vector is large enough,
3583 of course) are set to -1.
3585 Note: Elements in the first two-thirds of ovector that do not corre-
3586 spond to capturing parentheses in the pattern are never changed. That
3587 is, if a pattern contains n capturing parentheses, no more than ovec-
3588 tor[0] to ovector[2n+1] are set by pcre_exec(). The other elements (in
3589 the first two-thirds) retain whatever values they previously had.
3591 Some convenience functions are provided for extracting the captured
3592 substrings as separate strings. These are described below.
3594 Error return values from pcre_exec()
3596 If pcre_exec() fails, it returns a negative number. The following are
3597 defined in the header file:
3599 PCRE_ERROR_NOMATCH (-1)
3601 The subject string did not match the pattern.
3603 PCRE_ERROR_NULL (-2)
3605 Either code or subject was passed as NULL, or ovector was NULL and
3606 ovecsize was not zero.
3608 PCRE_ERROR_BADOPTION (-3)
3610 An unrecognized bit was set in the options argument.
3612 PCRE_ERROR_BADMAGIC (-4)
3614 PCRE stores a 4-byte "magic number" at the start of the compiled code,
3615 to catch the case when it is passed a junk pointer and to detect when a
3616 pattern that was compiled in an environment of one endianness is run in
3617 an environment with the other endianness. This is the error that PCRE
3618 gives when the magic number is not present.
3620 PCRE_ERROR_UNKNOWN_OPCODE (-5)
3622 While running the pattern match, an unknown item was encountered in the
3623 compiled pattern. This error could be caused by a bug in PCRE or by
3624 overwriting of the compiled pattern.
3626 PCRE_ERROR_NOMEMORY (-6)
3628 If a pattern contains back references, but the ovector that is passed
3629 to pcre_exec() is not big enough to remember the referenced substrings,
3630 PCRE gets a block of memory at the start of matching to use for this
3631 purpose. If the call via pcre_malloc() fails, this error is given. The
3632 memory is automatically freed at the end of matching.
3634 This error is also given if pcre_stack_malloc() fails in pcre_exec().
3635 This can happen only when PCRE has been compiled with --disable-stack-
3638 PCRE_ERROR_NOSUBSTRING (-7)
3640 This error is used by the pcre_copy_substring(), pcre_get_substring(),
3641 and pcre_get_substring_list() functions (see below). It is never
3642 returned by pcre_exec().
3644 PCRE_ERROR_MATCHLIMIT (-8)
3646 The backtracking limit, as specified by the match_limit field in a
3647 pcre_extra structure (or defaulted) was reached. See the description
3650 PCRE_ERROR_CALLOUT (-9)
3652 This error is never generated by pcre_exec() itself. It is provided for
3653 use by callout functions that want to yield a distinctive error code.
3654 See the pcrecallout documentation for details.
3656 PCRE_ERROR_BADUTF8 (-10)
3658 A string that contains an invalid UTF-8 byte sequence was passed as a
3659 subject, and the PCRE_NO_UTF8_CHECK option was not set. If the size of
3660 the output vector (ovecsize) is at least 2, the byte offset to the
3661 start of the the invalid UTF-8 character is placed in the first ele-
3662 ment, and a reason code is placed in the second element. The reason
3663 codes are listed in the following section. For backward compatibility,
3664 if PCRE_PARTIAL_HARD is set and the problem is a truncated UTF-8 char-
3665 acter at the end of the subject (reason codes 1 to 5),
3666 PCRE_ERROR_SHORTUTF8 is returned instead of PCRE_ERROR_BADUTF8.
3668 PCRE_ERROR_BADUTF8_OFFSET (-11)
3670 The UTF-8 byte sequence that was passed as a subject was checked and
3671 found to be valid (the PCRE_NO_UTF8_CHECK option was not set), but the
3672 value of startoffset did not point to the beginning of a UTF-8 charac-
3673 ter or the end of the subject.
3675 PCRE_ERROR_PARTIAL (-12)
3677 The subject string did not match, but it did match partially. See the
3678 pcrepartial documentation for details of partial matching.
3680 PCRE_ERROR_BADPARTIAL (-13)
3682 This code is no longer in use. It was formerly returned when the
3683 PCRE_PARTIAL option was used with a compiled pattern containing items
3684 that were not supported for partial matching. From release 8.00
3685 onwards, there are no restrictions on partial matching.
3687 PCRE_ERROR_INTERNAL (-14)
3689 An unexpected internal error has occurred. This error could be caused
3690 by a bug in PCRE or by overwriting of the compiled pattern.
3692 PCRE_ERROR_BADCOUNT (-15)
3694 This error is given if the value of the ovecsize argument is negative.
3696 PCRE_ERROR_RECURSIONLIMIT (-21)
3698 The internal recursion limit, as specified by the match_limit_recursion
3699 field in a pcre_extra structure (or defaulted) was reached. See the
3702 PCRE_ERROR_BADNEWLINE (-23)
3704 An invalid combination of PCRE_NEWLINE_xxx options was given.
3706 PCRE_ERROR_BADOFFSET (-24)
3708 The value of startoffset was negative or greater than the length of the
3709 subject, that is, the value in length.
3711 PCRE_ERROR_SHORTUTF8 (-25)
3713 This error is returned instead of PCRE_ERROR_BADUTF8 when the subject
3714 string ends with a truncated UTF-8 character and the PCRE_PARTIAL_HARD
3715 option is set. Information about the failure is returned as for
3716 PCRE_ERROR_BADUTF8. It is in fact sufficient to detect this case, but
3717 this special error code for PCRE_PARTIAL_HARD precedes the implementa-
3718 tion of returned information; it is retained for backwards compatibil-
3721 PCRE_ERROR_RECURSELOOP (-26)
3723 This error is returned when pcre_exec() detects a recursion loop within
3724 the pattern. Specifically, it means that either the whole pattern or a
3725 subpattern has been called recursively for the second time at the same
3726 position in the subject string. Some simple patterns that might do this
3727 are detected and faulted at compile time, but more complicated cases,
3728 in particular mutual recursions between two different subpatterns, can-
3729 not be detected until run time.
3731 PCRE_ERROR_JIT_STACKLIMIT (-27)
3733 This error is returned when a pattern that was successfully studied
3734 using a JIT compile option is being matched, but the memory available
3735 for the just-in-time processing stack is not large enough. See the
3736 pcrejit documentation for more details.
3738 PCRE_ERROR_BADMODE (-28)
3740 This error is given if a pattern that was compiled by the 8-bit library
3741 is passed to a 16-bit or 32-bit library function, or vice versa.
3743 PCRE_ERROR_BADENDIANNESS (-29)
3745 This error is given if a pattern that was compiled and saved is
3746 reloaded on a host with different endianness. The utility function
3747 pcre_pattern_to_host_byte_order() can be used to convert such a pattern
3748 so that it runs on the new host.
3750 PCRE_ERROR_JIT_BADOPTION
3752 This error is returned when a pattern that was successfully studied
3753 using a JIT compile option is being matched, but the matching mode
3754 (partial or complete match) does not correspond to any JIT compilation
3755 mode. When the JIT fast path function is used, this error may be also
3756 given for invalid options. See the pcrejit documentation for more
3759 PCRE_ERROR_BADLENGTH (-32)
3761 This error is given if pcre_exec() is called with a negative value for
3762 the length argument.
3764 Error numbers -16 to -20, -22, and 30 are not used by pcre_exec().
3766 Reason codes for invalid UTF-8 strings
3768 This section applies only to the 8-bit library. The corresponding
3769 information for the 16-bit and 32-bit libraries is given in the pcre16
3772 When pcre_exec() returns either PCRE_ERROR_BADUTF8 or PCRE_ERROR_SHORT-
3773 UTF8, and the size of the output vector (ovecsize) is at least 2, the
3774 offset of the start of the invalid UTF-8 character is placed in the
3775 first output vector element (ovector[0]) and a reason code is placed in
3776 the second element (ovector[1]). The reason codes are given names in
3777 the pcre.h header file:
3785 The string ends with a truncated UTF-8 character; the code specifies
3786 how many bytes are missing (1 to 5). Although RFC 3629 restricts UTF-8
3787 characters to be no longer than 4 bytes, the encoding scheme (origi-
3788 nally defined by RFC 2279) allows for up to 6 bytes, and this is
3789 checked first; hence the possibility of 4 or 5 missing bytes.
3797 The two most significant bits of the 2nd, 3rd, 4th, 5th, or 6th byte of
3798 the character do not have the binary value 0b10 (that is, either the
3799 most significant bit is 0, or the next bit is 1).
3804 A character that is valid by the RFC 2279 rules is either 5 or 6 bytes
3805 long; these code points are excluded by RFC 3629.
3809 A 4-byte character has a value greater than 0x10fff; these code points
3810 are excluded by RFC 3629.
3814 A 3-byte character has a value in the range 0xd800 to 0xdfff; this
3815 range of code points are reserved by RFC 3629 for use with UTF-16, and
3816 so are excluded from UTF-8.
3824 A 2-, 3-, 4-, 5-, or 6-byte character is "overlong", that is, it codes
3825 for a value that can be represented by fewer bytes, which is invalid.
3826 For example, the two bytes 0xc0, 0xae give the value 0x2e, whose cor-
3827 rect coding uses just one byte.
3831 The two most significant bits of the first byte of a character have the
3832 binary value 0b10 (that is, the most significant bit is 1 and the sec-
3833 ond is 0). Such a byte can only validly occur as the second or subse-
3834 quent byte of a multi-byte character.
3838 The first byte of a character has the value 0xfe or 0xff. These values
3839 can never occur in a valid UTF-8 string.
3843 This error code was formerly used when the presence of a so-called
3844 "non-character" caused an error. Unicode corrigendum #9 makes it clear
3845 that such characters should not cause a string to be rejected, and so
3846 this code is no longer in use and is never returned.
3849 EXTRACTING CAPTURED SUBSTRINGS BY NUMBER
3851 int pcre_copy_substring(const char *subject, int *ovector,
3852 int stringcount, int stringnumber, char *buffer,
3855 int pcre_get_substring(const char *subject, int *ovector,
3856 int stringcount, int stringnumber,
3857 const char **stringptr);
3859 int pcre_get_substring_list(const char *subject,
3860 int *ovector, int stringcount, const char ***listptr);
3862 Captured substrings can be accessed directly by using the offsets
3863 returned by pcre_exec() in ovector. For convenience, the functions
3864 pcre_copy_substring(), pcre_get_substring(), and pcre_get_sub-
3865 string_list() are provided for extracting captured substrings as new,
3866 separate, zero-terminated strings. These functions identify substrings
3867 by number. The next section describes functions for extracting named
3870 A substring that contains a binary zero is correctly extracted and has
3871 a further zero added on the end, but the result is not, of course, a C
3872 string. However, you can process such a string by referring to the
3873 length that is returned by pcre_copy_substring() and pcre_get_sub-
3874 string(). Unfortunately, the interface to pcre_get_substring_list() is
3875 not adequate for handling strings containing binary zeros, because the
3876 end of the final string is not independently indicated.
3878 The first three arguments are the same for all three of these func-
3879 tions: subject is the subject string that has just been successfully
3880 matched, ovector is a pointer to the vector of integer offsets that was
3881 passed to pcre_exec(), and stringcount is the number of substrings that
3882 were captured by the match, including the substring that matched the
3883 entire regular expression. This is the value returned by pcre_exec() if
3884 it is greater than zero. If pcre_exec() returned zero, indicating that
3885 it ran out of space in ovector, the value passed as stringcount should
3886 be the number of elements in the vector divided by three.
3888 The functions pcre_copy_substring() and pcre_get_substring() extract a
3889 single substring, whose number is given as stringnumber. A value of
3890 zero extracts the substring that matched the entire pattern, whereas
3891 higher values extract the captured substrings. For pcre_copy_sub-
3892 string(), the string is placed in buffer, whose length is given by
3893 buffersize, while for pcre_get_substring() a new block of memory is
3894 obtained via pcre_malloc, and its address is returned via stringptr.
3895 The yield of the function is the length of the string, not including
3896 the terminating zero, or one of these error codes:
3898 PCRE_ERROR_NOMEMORY (-6)
3900 The buffer was too small for pcre_copy_substring(), or the attempt to
3901 get memory failed for pcre_get_substring().
3903 PCRE_ERROR_NOSUBSTRING (-7)
3905 There is no substring whose number is stringnumber.
3907 The pcre_get_substring_list() function extracts all available sub-
3908 strings and builds a list of pointers to them. All this is done in a
3909 single block of memory that is obtained via pcre_malloc. The address of
3910 the memory block is returned via listptr, which is also the start of
3911 the list of string pointers. The end of the list is marked by a NULL
3912 pointer. The yield of the function is zero if all went well, or the
3915 PCRE_ERROR_NOMEMORY (-6)
3917 if the attempt to get the memory block failed.
3919 When any of these functions encounter a substring that is unset, which
3920 can happen when capturing subpattern number n+1 matches some part of
3921 the subject, but subpattern n has not been used at all, they return an
3922 empty string. This can be distinguished from a genuine zero-length sub-
3923 string by inspecting the appropriate offset in ovector, which is nega-
3924 tive for unset substrings.
3926 The two convenience functions pcre_free_substring() and pcre_free_sub-
3927 string_list() can be used to free the memory returned by a previous
3928 call of pcre_get_substring() or pcre_get_substring_list(), respec-
3929 tively. They do nothing more than call the function pointed to by
3930 pcre_free, which of course could be called directly from a C program.
3931 However, PCRE is used in some situations where it is linked via a spe-
3932 cial interface to another programming language that cannot use
3933 pcre_free directly; it is for these cases that the functions are pro-
3937 EXTRACTING CAPTURED SUBSTRINGS BY NAME
3939 int pcre_get_stringnumber(const pcre *code,
3942 int pcre_copy_named_substring(const pcre *code,
3943 const char *subject, int *ovector,
3944 int stringcount, const char *stringname,
3945 char *buffer, int buffersize);
3947 int pcre_get_named_substring(const pcre *code,
3948 const char *subject, int *ovector,
3949 int stringcount, const char *stringname,
3950 const char **stringptr);
3952 To extract a substring by name, you first have to find associated num-
3953 ber. For example, for this pattern
3957 the number of the subpattern called "xxx" is 2. If the name is known to
3958 be unique (PCRE_DUPNAMES was not set), you can find the number from the
3959 name by calling pcre_get_stringnumber(). The first argument is the com-
3960 piled pattern, and the second is the name. The yield of the function is
3961 the subpattern number, or PCRE_ERROR_NOSUBSTRING (-7) if there is no
3962 subpattern of that name.
3964 Given the number, you can extract the substring directly, or use one of
3965 the functions described in the previous section. For convenience, there
3966 are also two functions that do the whole job.
3968 Most of the arguments of pcre_copy_named_substring() and
3969 pcre_get_named_substring() are the same as those for the similarly
3970 named functions that extract by number. As these are described in the
3971 previous section, they are not re-described here. There are just two
3974 First, instead of a substring number, a substring name is given. Sec-
3975 ond, there is an extra argument, given at the start, which is a pointer
3976 to the compiled pattern. This is needed in order to gain access to the
3977 name-to-number translation table.
3979 These functions call pcre_get_stringnumber(), and if it succeeds, they
3980 then call pcre_copy_substring() or pcre_get_substring(), as appropri-
3981 ate. NOTE: If PCRE_DUPNAMES is set and there are duplicate names, the
3982 behaviour may not be what you want (see the next section).
3984 Warning: If the pattern uses the (?| feature to set up multiple subpat-
3985 terns with the same number, as described in the section on duplicate
3986 subpattern numbers in the pcrepattern page, you cannot use names to
3987 distinguish the different subpatterns, because names are not included
3988 in the compiled code. The matching process uses only numbers. For this
3989 reason, the use of different names for subpatterns of the same number
3990 causes an error at compile time.
3993 DUPLICATE SUBPATTERN NAMES
3995 int pcre_get_stringtable_entries(const pcre *code,
3996 const char *name, char **first, char **last);
3998 When a pattern is compiled with the PCRE_DUPNAMES option, names for
3999 subpatterns are not required to be unique. (Duplicate names are always
4000 allowed for subpatterns with the same number, created by using the (?|
4001 feature. Indeed, if such subpatterns are named, they are required to
4002 use the same names.)
4004 Normally, patterns with duplicate names are such that in any one match,
4005 only one of the named subpatterns participates. An example is shown in
4006 the pcrepattern documentation.
4008 When duplicates are present, pcre_copy_named_substring() and
4009 pcre_get_named_substring() return the first substring corresponding to
4010 the given name that is set. If none are set, PCRE_ERROR_NOSUBSTRING
4011 (-7) is returned; no data is returned. The pcre_get_stringnumber()
4012 function returns one of the numbers that are associated with the name,
4013 but it is not defined which it is.
4015 If you want to get full details of all captured substrings for a given
4016 name, you must use the pcre_get_stringtable_entries() function. The
4017 first argument is the compiled pattern, and the second is the name. The
4018 third and fourth are pointers to variables which are updated by the
4019 function. After it has run, they point to the first and last entries in
4020 the name-to-number table for the given name. The function itself
4021 returns the length of each entry, or PCRE_ERROR_NOSUBSTRING (-7) if
4022 there are none. The format of the table is described above in the sec-
4023 tion entitled Information about a pattern above. Given all the rele-
4024 vant entries for the name, you can extract each of their numbers, and
4025 hence the captured data, if any.
4028 FINDING ALL POSSIBLE MATCHES
4030 The traditional matching function uses a similar algorithm to Perl,
4031 which stops when it finds the first match, starting at a given point in
4032 the subject. If you want to find all possible matches, or the longest
4033 possible match, consider using the alternative matching function (see
4034 below) instead. If you cannot use the alternative function, but still
4035 need to find all possible matches, you can kludge it up by making use
4036 of the callout facility, which is described in the pcrecallout documen-
4039 What you have to do is to insert a callout right at the end of the pat-
4040 tern. When your callout function is called, extract and save the cur-
4041 rent matched substring. Then return 1, which forces pcre_exec() to
4042 backtrack and try other alternatives. Ultimately, when it runs out of
4043 matches, pcre_exec() will yield PCRE_ERROR_NOMATCH.
4046 OBTAINING AN ESTIMATE OF STACK USAGE
4048 Matching certain patterns using pcre_exec() can use a lot of process
4049 stack, which in certain environments can be rather limited in size.
4050 Some users find it helpful to have an estimate of the amount of stack
4051 that is used by pcre_exec(), to help them set recursion limits, as
4052 described in the pcrestack documentation. The estimate that is output
4053 by pcretest when called with the -m and -C options is obtained by call-
4054 ing pcre_exec with the values NULL, NULL, NULL, -999, and -999 for its
4055 first five arguments.
4057 Normally, if its first argument is NULL, pcre_exec() immediately
4058 returns the negative error code PCRE_ERROR_NULL, but with this special
4059 combination of arguments, it returns instead a negative number whose
4060 absolute value is the approximate stack frame size in bytes. (A nega-
4061 tive number is used so that it is clear that no match has happened.)
4062 The value is approximate because in some cases, recursive calls to
4063 pcre_exec() occur when there are one or two additional variables on the
4066 If PCRE has been compiled to use the heap instead of the stack for
4067 recursion, the value returned is the size of each block that is
4068 obtained from the heap.
4071 MATCHING A PATTERN: THE ALTERNATIVE FUNCTION
4073 int pcre_dfa_exec(const pcre *code, const pcre_extra *extra,
4074 const char *subject, int length, int startoffset,
4075 int options, int *ovector, int ovecsize,
4076 int *workspace, int wscount);
4078 The function pcre_dfa_exec() is called to match a subject string
4079 against a compiled pattern, using a matching algorithm that scans the
4080 subject string just once, and does not backtrack. This has different
4081 characteristics to the normal algorithm, and is not compatible with
4082 Perl. Some of the features of PCRE patterns are not supported. Never-
4083 theless, there are times when this kind of matching can be useful. For
4084 a discussion of the two matching algorithms, and a list of features
4085 that pcre_dfa_exec() does not support, see the pcrematching documenta-
4088 The arguments for the pcre_dfa_exec() function are the same as for
4089 pcre_exec(), plus two extras. The ovector argument is used in a differ-
4090 ent way, and this is described below. The other common arguments are
4091 used in the same way as for pcre_exec(), so their description is not
4094 The two additional arguments provide workspace for the function. The
4095 workspace vector should contain at least 20 elements. It is used for
4096 keeping track of multiple paths through the pattern tree. More
4097 workspace will be needed for patterns and subjects where there are a
4098 lot of potential matches.
4100 Here is an example of a simple call to pcre_dfa_exec():
4106 re, /* result of pcre_compile() */
4107 NULL, /* we didn't study the pattern */
4108 "some string", /* the subject string */
4109 11, /* the length of the subject string */
4110 0, /* start at offset 0 in the subject */
4111 0, /* default options */
4112 ovector, /* vector of integers for substring information */
4113 10, /* number of elements (NOT size in bytes) */
4114 wspace, /* working space vector */
4115 20); /* number of elements (NOT size in bytes) */
4117 Option bits for pcre_dfa_exec()
4119 The unused bits of the options argument for pcre_dfa_exec() must be
4120 zero. The only bits that may be set are PCRE_ANCHORED, PCRE_NEW-
4121 LINE_xxx, PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY,
4122 PCRE_NOTEMPTY_ATSTART, PCRE_NO_UTF8_CHECK, PCRE_BSR_ANYCRLF,
4123 PCRE_BSR_UNICODE, PCRE_NO_START_OPTIMIZE, PCRE_PARTIAL_HARD, PCRE_PAR-
4124 TIAL_SOFT, PCRE_DFA_SHORTEST, and PCRE_DFA_RESTART. All but the last
4125 four of these are exactly the same as for pcre_exec(), so their
4126 description is not repeated here.
4131 These have the same general effect as they do for pcre_exec(), but the
4132 details are slightly different. When PCRE_PARTIAL_HARD is set for
4133 pcre_dfa_exec(), it returns PCRE_ERROR_PARTIAL if the end of the sub-
4134 ject is reached and there is still at least one matching possibility
4135 that requires additional characters. This happens even if some complete
4136 matches have also been found. When PCRE_PARTIAL_SOFT is set, the return
4137 code PCRE_ERROR_NOMATCH is converted into PCRE_ERROR_PARTIAL if the end
4138 of the subject is reached, there have been no complete matches, but
4139 there is still at least one matching possibility. The portion of the
4140 string that was inspected when the longest partial match was found is
4141 set as the first matching string in both cases. There is a more
4142 detailed discussion of partial and multi-segment matching, with exam-
4143 ples, in the pcrepartial documentation.
4147 Setting the PCRE_DFA_SHORTEST option causes the matching algorithm to
4148 stop as soon as it has found one match. Because of the way the alterna-
4149 tive algorithm works, this is necessarily the shortest possible match
4150 at the first possible matching point in the subject string.
4154 When pcre_dfa_exec() returns a partial match, it is possible to call it
4155 again, with additional subject characters, and have it continue with
4156 the same match. The PCRE_DFA_RESTART option requests this action; when
4157 it is set, the workspace and wscount options must reference the same
4158 vector as before because data about the match so far is left in them
4159 after a partial match. There is more discussion of this facility in the
4160 pcrepartial documentation.
4162 Successful returns from pcre_dfa_exec()
4164 When pcre_dfa_exec() succeeds, it may have matched more than one sub-
4165 string in the subject. Note, however, that all the matches from one run
4166 of the function start at the same point in the subject. The shorter
4167 matches are all initial substrings of the longer matches. For example,
4172 is matched against the string
4174 This is <something> <something else> <something further> no more
4176 the three matched strings are
4179 <something> <something else>
4180 <something> <something else> <something further>
4182 On success, the yield of the function is a number greater than zero,
4183 which is the number of matched substrings. The substrings themselves
4184 are returned in ovector. Each string uses two elements; the first is
4185 the offset to the start, and the second is the offset to the end. In
4186 fact, all the strings have the same start offset. (Space could have
4187 been saved by giving this only once, but it was decided to retain some
4188 compatibility with the way pcre_exec() returns data, even though the
4189 meaning of the strings is different.)
4191 The strings are returned in reverse order of length; that is, the long-
4192 est matching string is given first. If there were too many matches to
4193 fit into ovector, the yield of the function is zero, and the vector is
4194 filled with the longest matches. Unlike pcre_exec(), pcre_dfa_exec()
4195 can use the entire ovector for returning matched strings.
4197 NOTE: PCRE's "auto-possessification" optimization usually applies to
4198 character repeats at the end of a pattern (as well as internally). For
4199 example, the pattern "a\d+" is compiled as if it were "a\d++" because
4200 there is no point even considering the possibility of backtracking into
4201 the repeated digits. For DFA matching, this means that only one possi-
4202 ble match is found. If you really do want multiple matches in such
4203 cases, either use an ungreedy repeat ("a\d+?") or set the
4204 PCRE_NO_AUTO_POSSESS option when compiling.
4206 Error returns from pcre_dfa_exec()
4208 The pcre_dfa_exec() function returns a negative number when it fails.
4209 Many of the errors are the same as for pcre_exec(), and these are
4210 described above. There are in addition the following errors that are
4211 specific to pcre_dfa_exec():
4213 PCRE_ERROR_DFA_UITEM (-16)
4215 This return is given if pcre_dfa_exec() encounters an item in the pat-
4216 tern that it does not support, for instance, the use of \C or a back
4219 PCRE_ERROR_DFA_UCOND (-17)
4221 This return is given if pcre_dfa_exec() encounters a condition item
4222 that uses a back reference for the condition, or a test for recursion
4223 in a specific group. These are not supported.
4225 PCRE_ERROR_DFA_UMLIMIT (-18)
4227 This return is given if pcre_dfa_exec() is called with an extra block
4228 that contains a setting of the match_limit or match_limit_recursion
4229 fields. This is not supported (these fields are meaningless for DFA
4232 PCRE_ERROR_DFA_WSSIZE (-19)
4234 This return is given if pcre_dfa_exec() runs out of space in the
4237 PCRE_ERROR_DFA_RECURSE (-20)
4239 When a recursive subpattern is processed, the matching function calls
4240 itself recursively, using private vectors for ovector and workspace.
4241 This error is given if the output vector is not large enough. This
4242 should be extremely rare, as a vector of size 1000 is used.
4244 PCRE_ERROR_DFA_BADRESTART (-30)
4246 When pcre_dfa_exec() is called with the PCRE_DFA_RESTART option, some
4247 plausibility checks are made on the contents of the workspace, which
4248 should contain data about the previous partial match. If any of these
4249 checks fail, this error is given.
4254 pcre16(3), pcre32(3), pcrebuild(3), pcrecallout(3), pcrecpp(3)(3),
4255 pcrematching(3), pcrepartial(3), pcreposix(3), pcreprecompile(3), pcre-
4256 sample(3), pcrestack(3).
4262 University Computing Service
4263 Cambridge CB2 3QH, England.
4268 Last updated: 09 February 2014
4269 Copyright (c) 1997-2014 University of Cambridge.
4270 ------------------------------------------------------------------------------
4273 PCRECALLOUT(3) Library Functions Manual PCRECALLOUT(3)
4278 PCRE - Perl-compatible regular expressions
4284 int (*pcre_callout)(pcre_callout_block *);
4286 int (*pcre16_callout)(pcre16_callout_block *);
4288 int (*pcre32_callout)(pcre32_callout_block *);
4293 PCRE provides a feature called "callout", which is a means of temporar-
4294 ily passing control to the caller of PCRE in the middle of pattern
4295 matching. The caller of PCRE provides an external function by putting
4296 its entry point in the global variable pcre_callout (pcre16_callout for
4297 the 16-bit library, pcre32_callout for the 32-bit library). By default,
4298 this variable contains NULL, which disables all calling out.
4300 Within a regular expression, (?C) indicates the points at which the
4301 external function is to be called. Different callout points can be
4302 identified by putting a number less than 256 after the letter C. The
4303 default value is zero. For example, this pattern has two callout
4308 If the PCRE_AUTO_CALLOUT option bit is set when a pattern is compiled,
4309 PCRE automatically inserts callouts, all with number 255, before each
4310 item in the pattern. For example, if PCRE_AUTO_CALLOUT is used with the
4315 it is processed as if it were
4317 (?C255)A(?C255)((?C255)\d{2}(?C255)|(?C255)-(?C255)-(?C255))(?C255)
4319 Notice that there is a callout before and after each parenthesis and
4320 alternation bar. If the pattern contains a conditional group whose con-
4321 dition is an assertion, an automatic callout is inserted immediately
4322 before the condition. Such a callout may also be inserted explicitly,
4327 This applies only to assertion conditions (because they are themselves
4328 independent groups).
4330 Automatic callouts can be used for tracking the progress of pattern
4331 matching. The pcretest program has a pattern qualifier (/C) that sets
4332 automatic callouts; when it is used, the output indicates how the pat-
4333 tern is being matched. This is useful information when you are trying
4334 to optimize the performance of a particular pattern.
4339 You should be aware that, because of optimizations in the way PCRE com-
4340 piles and matches patterns, callouts sometimes do not happen exactly as
4343 At compile time, PCRE "auto-possessifies" repeated items when it knows
4344 that what follows cannot be part of the repeat. For example, a+[bc] is
4345 compiled as if it were a++[bc]. The pcretest output when this pattern
4346 is anchored and then applied with automatic callouts to the string
4355 This indicates that when matching [bc] fails, there is no backtracking
4356 into a+ and therefore the callouts that would be taken for the back-
4357 tracks do not occur. You can disable the auto-possessify feature by
4358 passing PCRE_NO_AUTO_POSSESS to pcre_compile(), or starting the pattern
4359 with (*NO_AUTO_POSSESS). If this is done in pcretest (using the /O
4360 qualifier), the output changes to this:
4371 This time, when matching [bc] fails, the matcher backtracks into a+ and
4372 tries again, repeatedly, until a+ itself fails.
4374 Other optimizations that provide fast "no match" results also affect
4375 callouts. For example, if the pattern is
4379 PCRE knows that any matching string must contain the letter "d". If the
4380 subject string is "abyz", the lack of "d" means that matching doesn't
4381 ever start, and the callout is never reached. However, with "abyd",
4382 though the result is still no match, the callout is obeyed.
4384 If the pattern is studied, PCRE knows the minimum length of a matching
4385 string, and will immediately give a "no match" return without actually
4386 running a match if the subject is not long enough, or, for unanchored
4387 patterns, if it has been scanned far enough.
4389 You can disable these optimizations by passing the PCRE_NO_START_OPTI-
4390 MIZE option to the matching function, or by starting the pattern with
4391 (*NO_START_OPT). This slows down the matching process, but does ensure
4392 that callouts such as the example above are obeyed.
4395 THE CALLOUT INTERFACE
4397 During matching, when PCRE reaches a callout point, the external func-
4398 tion defined by pcre_callout or pcre[16|32]_callout is called (if it is
4399 set). This applies to both normal and DFA matching. The only argument
4400 to the callout function is a pointer to a pcre_callout or
4401 pcre[16|32]_callout block. These structures contains the following
4407 const char *subject; (8-bit version)
4408 PCRE_SPTR16 subject; (16-bit version)
4409 PCRE_SPTR32 subject; (32-bit version)
4412 int current_position;
4416 int pattern_position;
4417 int next_item_length;
4418 const unsigned char *mark; (8-bit version)
4419 const PCRE_UCHAR16 *mark; (16-bit version)
4420 const PCRE_UCHAR32 *mark; (32-bit version)
4422 The version field is an integer containing the version number of the
4423 block format. The initial version was 0; the current version is 2. The
4424 version number will change again in future if additional fields are
4425 added, but the intention is never to remove any of the existing fields.
4427 The callout_number field contains the number of the callout, as com-
4428 piled into the pattern (that is, the number after ?C for manual call-
4429 outs, and 255 for automatically generated callouts).
4431 The offset_vector field is a pointer to the vector of offsets that was
4432 passed by the caller to the matching function. When pcre_exec() or
4433 pcre[16|32]_exec() is used, the contents can be inspected, in order to
4434 extract substrings that have been matched so far, in the same way as
4435 for extracting substrings after a match has completed. For the DFA
4436 matching functions, this field is not useful.
4438 The subject and subject_length fields contain copies of the values that
4439 were passed to the matching function.
4441 The start_match field normally contains the offset within the subject
4442 at which the current match attempt started. However, if the escape
4443 sequence \K has been encountered, this value is changed to reflect the
4444 modified starting point. If the pattern is not anchored, the callout
4445 function may be called several times from the same point in the pattern
4446 for different starting points in the subject.
4448 The current_position field contains the offset within the subject of
4449 the current match pointer.
4451 When the pcre_exec() or pcre[16|32]_exec() is used, the capture_top
4452 field contains one more than the number of the highest numbered cap-
4453 tured substring so far. If no substrings have been captured, the value
4454 of capture_top is one. This is always the case when the DFA functions
4455 are used, because they do not support captured substrings.
4457 The capture_last field contains the number of the most recently cap-
4458 tured substring. However, when a recursion exits, the value reverts to
4459 what it was outside the recursion, as do the values of all captured
4460 substrings. If no substrings have been captured, the value of cap-
4461 ture_last is -1. This is always the case for the DFA matching func-
4464 The callout_data field contains a value that is passed to a matching
4465 function specifically so that it can be passed back in callouts. It is
4466 passed in the callout_data field of a pcre_extra or pcre[16|32]_extra
4467 data structure. If no such data was passed, the value of callout_data
4468 in a callout block is NULL. There is a description of the pcre_extra
4469 structure in the pcreapi documentation.
4471 The pattern_position field is present from version 1 of the callout
4472 structure. It contains the offset to the next item to be matched in the
4475 The next_item_length field is present from version 1 of the callout
4476 structure. It contains the length of the next item to be matched in the
4477 pattern string. When the callout immediately precedes an alternation
4478 bar, a closing parenthesis, or the end of the pattern, the length is
4479 zero. When the callout precedes an opening parenthesis, the length is
4480 that of the entire subpattern.
4482 The pattern_position and next_item_length fields are intended to help
4483 in distinguishing between different automatic callouts, which all have
4484 the same callout number. However, they are set for all callouts.
4486 The mark field is present from version 2 of the callout structure. In
4487 callouts from pcre_exec() or pcre[16|32]_exec() it contains a pointer
4488 to the zero-terminated name of the most recently passed (*MARK),
4489 (*PRUNE), or (*THEN) item in the match, or NULL if no such items have
4490 been passed. Instances of (*PRUNE) or (*THEN) without a name do not
4491 obliterate a previous (*MARK). In callouts from the DFA matching func-
4492 tions this field always contains NULL.
4497 The external callout function returns an integer to PCRE. If the value
4498 is zero, matching proceeds as normal. If the value is greater than
4499 zero, matching fails at the current point, but the testing of other
4500 matching possibilities goes ahead, just as if a lookahead assertion had
4501 failed. If the value is less than zero, the match is abandoned, the
4502 matching function returns the negative value.
4504 Negative values should normally be chosen from the set of
4505 PCRE_ERROR_xxx values. In particular, PCRE_ERROR_NOMATCH forces a stan-
4506 dard "no match" failure. The error number PCRE_ERROR_CALLOUT is
4507 reserved for use by callout functions; it will never be used by PCRE
4514 University Computing Service
4515 Cambridge CB2 3QH, England.
4520 Last updated: 12 November 2013
4521 Copyright (c) 1997-2013 University of Cambridge.
4522 ------------------------------------------------------------------------------
4525 PCRECOMPAT(3) Library Functions Manual PCRECOMPAT(3)
4530 PCRE - Perl-compatible regular expressions
4532 DIFFERENCES BETWEEN PCRE AND PERL
4534 This document describes the differences in the ways that PCRE and Perl
4535 handle regular expressions. The differences described here are with
4536 respect to Perl versions 5.10 and above.
4538 1. PCRE has only a subset of Perl's Unicode support. Details of what it
4539 does have are given in the pcreunicode page.
4541 2. PCRE allows repeat quantifiers only on parenthesized assertions, but
4542 they do not mean what you might think. For example, (?!a){3} does not
4543 assert that the next three characters are not "a". It just asserts that
4544 the next character is not "a" three times (in principle: PCRE optimizes
4545 this to run the assertion just once). Perl allows repeat quantifiers on
4546 other assertions such as \b, but these do not seem to have any use.
4548 3. Capturing subpatterns that occur inside negative lookahead asser-
4549 tions are counted, but their entries in the offsets vector are never
4550 set. Perl sometimes (but not always) sets its numerical variables from
4551 inside negative assertions.
4553 4. Though binary zero characters are supported in the subject string,
4554 they are not allowed in a pattern string because it is passed as a nor-
4555 mal C string, terminated by zero. The escape sequence \0 can be used in
4556 the pattern to represent a binary zero.
4558 5. The following Perl escape sequences are not supported: \l, \u, \L,
4559 \U, and \N when followed by a character name or Unicode value. (\N on
4560 its own, matching a non-newline character, is supported.) In fact these
4561 are implemented by Perl's general string-handling and are not part of
4562 its pattern matching engine. If any of these are encountered by PCRE,
4563 an error is generated by default. However, if the PCRE_JAVASCRIPT_COM-
4564 PAT option is set, \U and \u are interpreted as JavaScript interprets
4567 6. The Perl escape sequences \p, \P, and \X are supported only if PCRE
4568 is built with Unicode character property support. The properties that
4569 can be tested with \p and \P are limited to the general category prop-
4570 erties such as Lu and Nd, script names such as Greek or Han, and the
4571 derived properties Any and L&. PCRE does support the Cs (surrogate)
4572 property, which Perl does not; the Perl documentation says "Because
4573 Perl hides the need for the user to understand the internal representa-
4574 tion of Unicode characters, there is no need to implement the somewhat
4575 messy concept of surrogates."
4577 7. PCRE does support the \Q...\E escape for quoting substrings. Charac-
4578 ters in between are treated as literals. This is slightly different
4579 from Perl in that $ and @ are also handled as literals inside the
4580 quotes. In Perl, they cause variable interpolation (but of course PCRE
4581 does not have variables). Note the following examples:
4583 Pattern PCRE matches Perl matches
4585 \Qabc$xyz\E abc$xyz abc followed by the
4587 \Qabc\$xyz\E abc\$xyz abc\$xyz
4588 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
4590 The \Q...\E sequence is recognized both inside and outside character
4593 8. Fairly obviously, PCRE does not support the (?{code}) and (??{code})
4594 constructions. However, there is support for recursive patterns. This
4595 is not available in Perl 5.8, but it is in Perl 5.10. Also, the PCRE
4596 "callout" feature allows an external function to be called during pat-
4597 tern matching. See the pcrecallout documentation for details.
4599 9. Subpatterns that are called as subroutines (whether or not recur-
4600 sively) are always treated as atomic groups in PCRE. This is like
4601 Python, but unlike Perl. Captured values that are set outside a sub-
4602 routine call can be reference from inside in PCRE, but not in Perl.
4603 There is a discussion that explains these differences in more detail in
4604 the section on recursion differences from Perl in the pcrepattern page.
4606 10. If any of the backtracking control verbs are used in a subpattern
4607 that is called as a subroutine (whether or not recursively), their
4608 effect is confined to that subpattern; it does not extend to the sur-
4609 rounding pattern. This is not always the case in Perl. In particular,
4610 if (*THEN) is present in a group that is called as a subroutine, its
4611 action is limited to that group, even if the group does not contain any
4612 | characters. Note that such subpatterns are processed as anchored at
4613 the point where they are tested.
4615 11. If a pattern contains more than one backtracking control verb, the
4616 first one that is backtracked onto acts. For example, in the pattern
4617 A(*COMMIT)B(*PRUNE)C a failure in B triggers (*COMMIT), but a failure
4618 in C triggers (*PRUNE). Perl's behaviour is more complex; in many cases
4619 it is the same as PCRE, but there are examples where it differs.
4621 12. Most backtracking verbs in assertions have their normal actions.
4622 They are not confined to the assertion.
4624 13. There are some differences that are concerned with the settings of
4625 captured strings when part of a pattern is repeated. For example,
4626 matching "aba" against the pattern /^(a(b)?)+$/ in Perl leaves $2
4627 unset, but in PCRE it is set to "b".
4629 14. PCRE's handling of duplicate subpattern numbers and duplicate sub-
4630 pattern names is not as general as Perl's. This is a consequence of the
4631 fact the PCRE works internally just with numbers, using an external ta-
4632 ble to translate between numbers and names. In particular, a pattern
4633 such as (?|(?<a>A)|(?<b)B), where the two capturing parentheses have
4634 the same number but different names, is not supported, and causes an
4635 error at compile time. If it were allowed, it would not be possible to
4636 distinguish which parentheses matched, because both names map to cap-
4637 turing subpattern number 1. To avoid this confusing situation, an error
4638 is given at compile time.
4640 15. Perl recognizes comments in some places that PCRE does not, for
4641 example, between the ( and ? at the start of a subpattern. If the /x
4642 modifier is set, Perl allows white space between ( and ? (though cur-
4643 rent Perls warn that this is deprecated) but PCRE never does, even if
4644 the PCRE_EXTENDED option is set.
4646 16. Perl, when in warning mode, gives warnings for character classes
4647 such as [A-\d] or [a-[:digit:]]. It then treats the hyphens as liter-
4648 als. PCRE has no warning features, so it gives an error in these cases
4649 because they are almost certainly user mistakes.
4651 17. In PCRE, the upper/lower case character properties Lu and Ll are
4652 not affected when case-independent matching is specified. For example,
4653 \p{Lu} always matches an upper case letter. I think Perl has changed in
4654 this respect; in the release at the time of writing (5.16), \p{Lu} and
4655 \p{Ll} match all letters, regardless of case, when case independence is
4658 18. PCRE provides some extensions to the Perl regular expression facil-
4659 ities. Perl 5.10 includes new features that are not in earlier ver-
4660 sions of Perl, some of which (such as named parentheses) have been in
4661 PCRE for some time. This list is with respect to Perl 5.10:
4663 (a) Although lookbehind assertions in PCRE must match fixed length
4664 strings, each alternative branch of a lookbehind assertion can match a
4665 different length of string. Perl requires them all to have the same
4668 (b) If PCRE_DOLLAR_ENDONLY is set and PCRE_MULTILINE is not set, the $
4669 meta-character matches only at the very end of the string.
4671 (c) If PCRE_EXTRA is set, a backslash followed by a letter with no spe-
4672 cial meaning is faulted. Otherwise, like Perl, the backslash is quietly
4673 ignored. (Perl can be made to issue a warning.)
4675 (d) If PCRE_UNGREEDY is set, the greediness of the repetition quanti-
4676 fiers is inverted, that is, by default they are not greedy, but if fol-
4677 lowed by a question mark they are.
4679 (e) PCRE_ANCHORED can be used at matching time to force a pattern to be
4680 tried only at the first matching position in the subject string.
4682 (f) The PCRE_NOTBOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART,
4683 and PCRE_NO_AUTO_CAPTURE options for pcre_exec() have no Perl equiva-
4686 (g) The \R escape sequence can be restricted to match only CR, LF, or
4687 CRLF by the PCRE_BSR_ANYCRLF option.
4689 (h) The callout facility is PCRE-specific.
4691 (i) The partial matching facility is PCRE-specific.
4693 (j) Patterns compiled by PCRE can be saved and re-used at a later time,
4694 even on different hosts that have the other endianness. However, this
4695 does not apply to optimized data created by the just-in-time compiler.
4697 (k) The alternative matching functions (pcre_dfa_exec(),
4698 pcre16_dfa_exec() and pcre32_dfa_exec(),) match in a different way and
4699 are not Perl-compatible.
4701 (l) PCRE recognizes some special sequences such as (*CR) at the start
4702 of a pattern that set overall options that cannot be changed within the
4709 University Computing Service
4710 Cambridge CB2 3QH, England.
4715 Last updated: 10 November 2013
4716 Copyright (c) 1997-2013 University of Cambridge.
4717 ------------------------------------------------------------------------------
4720 PCREPATTERN(3) Library Functions Manual PCREPATTERN(3)
4725 PCRE - Perl-compatible regular expressions
4727 PCRE REGULAR EXPRESSION DETAILS
4729 The syntax and semantics of the regular expressions that are supported
4730 by PCRE are described in detail below. There is a quick-reference syn-
4731 tax summary in the pcresyntax page. PCRE tries to match Perl syntax and
4732 semantics as closely as it can. PCRE also supports some alternative
4733 regular expression syntax (which does not conflict with the Perl syn-
4734 tax) in order to provide some compatibility with regular expressions in
4735 Python, .NET, and Oniguruma.
4737 Perl's regular expressions are described in its own documentation, and
4738 regular expressions in general are covered in a number of books, some
4739 of which have copious examples. Jeffrey Friedl's "Mastering Regular
4740 Expressions", published by O'Reilly, covers regular expressions in
4741 great detail. This description of PCRE's regular expressions is
4742 intended as reference material.
4744 This document discusses the patterns that are supported by PCRE when
4745 one its main matching functions, pcre_exec() (8-bit) or
4746 pcre[16|32]_exec() (16- or 32-bit), is used. PCRE also has alternative
4747 matching functions, pcre_dfa_exec() and pcre[16|32_dfa_exec(), which
4748 match using a different algorithm that is not Perl-compatible. Some of
4749 the features discussed below are not available when DFA matching is
4750 used. The advantages and disadvantages of the alternative functions,
4751 and how they differ from the normal functions, are discussed in the
4755 SPECIAL START-OF-PATTERN ITEMS
4757 A number of options that can be passed to pcre_compile() can also be
4758 set by special items at the start of a pattern. These are not Perl-com-
4759 patible, but are provided to make these options accessible to pattern
4760 writers who are not able to change the program that processes the pat-
4761 tern. Any number of these items may appear, but they must all be
4762 together right at the start of the pattern string, and the letters must
4767 The original operation of PCRE was on strings of one-byte characters.
4768 However, there is now also support for UTF-8 strings in the original
4769 library, an extra library that supports 16-bit and UTF-16 character
4770 strings, and a third library that supports 32-bit and UTF-32 character
4771 strings. To use these features, PCRE must be built to include appropri-
4772 ate support. When using UTF strings you must either call the compiling
4773 function with the PCRE_UTF8, PCRE_UTF16, or PCRE_UTF32 option, or the
4774 pattern must start with one of these special sequences:
4781 (*UTF) is a generic sequence that can be used with any of the
4782 libraries. Starting a pattern with such a sequence is equivalent to
4783 setting the relevant option. How setting a UTF mode affects pattern
4784 matching is mentioned in several places below. There is also a summary
4785 of features in the pcreunicode page.
4787 Some applications that allow their users to supply patterns may wish to
4788 restrict them to non-UTF data for security reasons. If the
4789 PCRE_NEVER_UTF option is set at compile time, (*UTF) etc. are not
4790 allowed, and their appearance causes an error.
4792 Unicode property support
4794 Another special sequence that may appear at the start of a pattern is
4795 (*UCP). This has the same effect as setting the PCRE_UCP option: it
4796 causes sequences such as \d and \w to use Unicode properties to deter-
4797 mine character types, instead of recognizing only characters with codes
4798 less than 128 via a lookup table.
4800 Disabling auto-possessification
4802 If a pattern starts with (*NO_AUTO_POSSESS), it has the same effect as
4803 setting the PCRE_NO_AUTO_POSSESS option at compile time. This stops
4804 PCRE from making quantifiers possessive when what follows cannot match
4805 the repeated item. For example, by default a+b is treated as a++b. For
4806 more details, see the pcreapi documentation.
4808 Disabling start-up optimizations
4810 If a pattern starts with (*NO_START_OPT), it has the same effect as
4811 setting the PCRE_NO_START_OPTIMIZE option either at compile or matching
4812 time. This disables several optimizations for quickly reaching "no
4813 match" results. For more details, see the pcreapi documentation.
4817 PCRE supports five different conventions for indicating line breaks in
4818 strings: a single CR (carriage return) character, a single LF (line-
4819 feed) character, the two-character sequence CRLF, any of the three pre-
4820 ceding, or any Unicode newline sequence. The pcreapi page has further
4821 discussion about newlines, and shows how to set the newline convention
4822 in the options arguments for the compiling and matching functions.
4824 It is also possible to specify a newline convention by starting a pat-
4825 tern string with one of the following five sequences:
4827 (*CR) carriage return
4829 (*CRLF) carriage return, followed by linefeed
4830 (*ANYCRLF) any of the three above
4831 (*ANY) all Unicode newline sequences
4833 These override the default and the options given to the compiling func-
4834 tion. For example, on a Unix system where LF is the default newline
4835 sequence, the pattern
4839 changes the convention to CR. That pattern matches "a\nb" because LF is
4840 no longer a newline. If more than one of these settings is present, the
4843 The newline convention affects where the circumflex and dollar asser-
4844 tions are true. It also affects the interpretation of the dot metachar-
4845 acter when PCRE_DOTALL is not set, and the behaviour of \N. However, it
4846 does not affect what the \R escape sequence matches. By default, this
4847 is any Unicode newline sequence, for Perl compatibility. However, this
4848 can be changed; see the description of \R in the section entitled "New-
4849 line sequences" below. A change of \R setting can be combined with a
4850 change of newline convention.
4852 Setting match and recursion limits
4854 The caller of pcre_exec() can set a limit on the number of times the
4855 internal match() function is called and on the maximum depth of recur-
4856 sive calls. These facilities are provided to catch runaway matches that
4857 are provoked by patterns with huge matching trees (a typical example is
4858 a pattern with nested unlimited repeats) and to avoid running out of
4859 system stack by too much recursion. When one of these limits is
4860 reached, pcre_exec() gives an error return. The limits can also be set
4861 by items at the start of the pattern of the form
4864 (*LIMIT_RECURSION=d)
4866 where d is any number of decimal digits. However, the value of the set-
4867 ting must be less than the value set (or defaulted) by the caller of
4868 pcre_exec() for it to have any effect. In other words, the pattern
4869 writer can lower the limits set by the programmer, but not raise them.
4870 If there is more than one setting of one of these limits, the lower
4874 EBCDIC CHARACTER CODES
4876 PCRE can be compiled to run in an environment that uses EBCDIC as its
4877 character code rather than ASCII or Unicode (typically a mainframe sys-
4878 tem). In the sections below, character code values are ASCII or Uni-
4879 code; in an EBCDIC environment these characters may have different code
4880 values, and there are no code points greater than 255.
4883 CHARACTERS AND METACHARACTERS
4885 A regular expression is a pattern that is matched against a subject
4886 string from left to right. Most characters stand for themselves in a
4887 pattern, and match the corresponding characters in the subject. As a
4888 trivial example, the pattern
4892 matches a portion of a subject string that is identical to itself. When
4893 caseless matching is specified (the PCRE_CASELESS option), letters are
4894 matched independently of case. In a UTF mode, PCRE always understands
4895 the concept of case for characters whose values are less than 128, so
4896 caseless matching is always possible. For characters with higher val-
4897 ues, the concept of case is supported if PCRE is compiled with Unicode
4898 property support, but not otherwise. If you want to use caseless
4899 matching for characters 128 and above, you must ensure that PCRE is
4900 compiled with Unicode property support as well as with UTF support.
4902 The power of regular expressions comes from the ability to include
4903 alternatives and repetitions in the pattern. These are encoded in the
4904 pattern by the use of metacharacters, which do not stand for themselves
4905 but instead are interpreted in some special way.
4907 There are two different sets of metacharacters: those that are recog-
4908 nized anywhere in the pattern except within square brackets, and those
4909 that are recognized within square brackets. Outside square brackets,
4910 the metacharacters are as follows:
4912 \ general escape character with several uses
4913 ^ assert start of string (or line, in multiline mode)
4914 $ assert end of string (or line, in multiline mode)
4915 . match any character except newline (by default)
4916 [ start character class definition
4917 | start of alternative branch
4920 ? extends the meaning of (
4921 also 0 or 1 quantifier
4922 also quantifier minimizer
4923 * 0 or more quantifier
4924 + 1 or more quantifier
4925 also "possessive quantifier"
4926 { start min/max quantifier
4928 Part of a pattern that is in square brackets is called a "character
4929 class". In a character class the only metacharacters are:
4931 \ general escape character
4932 ^ negate the class, but only if the first character
4933 - indicates character range
4934 [ POSIX character class (only if followed by POSIX
4936 ] terminates the character class
4938 The following sections describe the use of each of the metacharacters.
4943 The backslash character has several uses. Firstly, if it is followed by
4944 a character that is not a number or a letter, it takes away any special
4945 meaning that character may have. This use of backslash as an escape
4946 character applies both inside and outside character classes.
4948 For example, if you want to match a * character, you write \* in the
4949 pattern. This escaping action applies whether or not the following
4950 character would otherwise be interpreted as a metacharacter, so it is
4951 always safe to precede a non-alphanumeric with backslash to specify
4952 that it stands for itself. In particular, if you want to match a back-
4953 slash, you write \\.
4955 In a UTF mode, only ASCII numbers and letters have any special meaning
4956 after a backslash. All other characters (in particular, those whose
4957 codepoints are greater than 127) are treated as literals.
4959 If a pattern is compiled with the PCRE_EXTENDED option, most white
4960 space in the pattern (other than in a character class), and characters
4961 between a # outside a character class and the next newline, inclusive,
4962 are ignored. An escaping backslash can be used to include a white space
4963 or # character as part of the pattern.
4965 If you want to remove the special meaning from a sequence of charac-
4966 ters, you can do so by putting them between \Q and \E. This is differ-
4967 ent from Perl in that $ and @ are handled as literals in \Q...\E
4968 sequences in PCRE, whereas in Perl, $ and @ cause variable interpola-
4969 tion. Note the following examples:
4971 Pattern PCRE matches Perl matches
4973 \Qabc$xyz\E abc$xyz abc followed by the
4975 \Qabc\$xyz\E abc\$xyz abc\$xyz
4976 \Qabc\E\$\Qxyz\E abc$xyz abc$xyz
4978 The \Q...\E sequence is recognized both inside and outside character
4979 classes. An isolated \E that is not preceded by \Q is ignored. If \Q
4980 is not followed by \E later in the pattern, the literal interpretation
4981 continues to the end of the pattern (that is, \E is assumed at the
4982 end). If the isolated \Q is inside a character class, this causes an
4983 error, because the character class is not terminated.
4985 Non-printing characters
4987 A second use of backslash provides a way of encoding non-printing char-
4988 acters in patterns in a visible manner. There is no restriction on the
4989 appearance of non-printing characters, apart from the binary zero that
4990 terminates a pattern, but when a pattern is being prepared by text
4991 editing, it is often easier to use one of the following escape
4992 sequences than the binary character it represents:
4994 \a alarm, that is, the BEL character (hex 07)
4995 \cx "control-x", where x is any ASCII character
4997 \f form feed (hex 0C)
4998 \n linefeed (hex 0A)
4999 \r carriage return (hex 0D)
5001 \0dd character with octal code 0dd
5002 \ddd character with octal code ddd, or back reference
5003 \o{ddd..} character with octal code ddd..
5004 \xhh character with hex code hh
5005 \x{hhh..} character with hex code hhh.. (non-JavaScript mode)
5006 \uhhhh character with hex code hhhh (JavaScript mode only)
5008 The precise effect of \cx on ASCII characters is as follows: if x is a
5009 lower case letter, it is converted to upper case. Then bit 6 of the
5010 character (hex 40) is inverted. Thus \cA to \cZ become hex 01 to hex 1A
5011 (A is 41, Z is 5A), but \c{ becomes hex 3B ({ is 7B), and \c; becomes
5012 hex 7B (; is 3B). If the data item (byte or 16-bit value) following \c
5013 has a value greater than 127, a compile-time error occurs. This locks
5014 out non-ASCII characters in all modes.
5016 The \c facility was designed for use with ASCII characters, but with
5017 the extension to Unicode it is even less useful than it once was. It
5018 is, however, recognized when PCRE is compiled in EBCDIC mode, where
5019 data items are always bytes. In this mode, all values are valid after
5020 \c. If the next character is a lower case letter, it is converted to
5021 upper case. Then the 0xc0 bits of the byte are inverted. Thus \cA
5022 becomes hex 01, as in ASCII (A is C1), but because the EBCDIC letters
5023 are disjoint, \cZ becomes hex 29 (Z is E9), and other characters also
5024 generate different values.
5026 After \0 up to two further octal digits are read. If there are fewer
5027 than two digits, just those that are present are used. Thus the
5028 sequence \0\x\07 specifies two binary zeros followed by a BEL character
5029 (code value 7). Make sure you supply two digits after the initial zero
5030 if the pattern character that follows is itself an octal digit.
5032 The escape \o must be followed by a sequence of octal digits, enclosed
5033 in braces. An error occurs if this is not the case. This escape is a
5034 recent addition to Perl; it provides way of specifying character code
5035 points as octal numbers greater than 0777, and it also allows octal
5036 numbers and back references to be unambiguously specified.
5038 For greater clarity and unambiguity, it is best to avoid following \ by
5039 a digit greater than zero. Instead, use \o{} or \x{} to specify charac-
5040 ter numbers, and \g{} to specify back references. The following para-
5041 graphs describe the old, ambiguous syntax.
5043 The handling of a backslash followed by a digit other than 0 is compli-
5044 cated, and Perl has changed in recent releases, causing PCRE also to
5045 change. Outside a character class, PCRE reads the digit and any follow-
5046 ing digits as a decimal number. If the number is less than 8, or if
5047 there have been at least that many previous capturing left parentheses
5048 in the expression, the entire sequence is taken as a back reference. A
5049 description of how this works is given later, following the discussion
5050 of parenthesized subpatterns.
5052 Inside a character class, or if the decimal number following \ is
5053 greater than 7 and there have not been that many capturing subpatterns,
5054 PCRE handles \8 and \9 as the literal characters "8" and "9", and oth-
5055 erwise re-reads up to three octal digits following the backslash, using
5056 them to generate a data character. Any subsequent digits stand for
5057 themselves. For example:
5059 \040 is another way of writing an ASCII space
5060 \40 is the same, provided there are fewer than 40
5061 previous capturing subpatterns
5062 \7 is always a back reference
5063 \11 might be a back reference, or another way of
5065 \011 is always a tab
5066 \0113 is a tab followed by the character "3"
5067 \113 might be a back reference, otherwise the
5068 character with octal code 113
5069 \377 might be a back reference, otherwise
5070 the value 255 (decimal)
5071 \81 is either a back reference, or the two
5072 characters "8" and "1"
5074 Note that octal values of 100 or greater that are specified using this
5075 syntax must not be introduced by a leading zero, because no more than
5076 three octal digits are ever read.
5078 By default, after \x that is not followed by {, from zero to two hexa-
5079 decimal digits are read (letters can be in upper or lower case). Any
5080 number of hexadecimal digits may appear between \x{ and }. If a charac-
5081 ter other than a hexadecimal digit appears between \x{ and }, or if
5082 there is no terminating }, an error occurs.
5084 If the PCRE_JAVASCRIPT_COMPAT option is set, the interpretation of \x
5085 is as just described only when it is followed by two hexadecimal dig-
5086 its. Otherwise, it matches a literal "x" character. In JavaScript
5087 mode, support for code points greater than 256 is provided by \u, which
5088 must be followed by four hexadecimal digits; otherwise it matches a
5089 literal "u" character.
5091 Characters whose value is less than 256 can be defined by either of the
5092 two syntaxes for \x (or by \u in JavaScript mode). There is no differ-
5093 ence in the way they are handled. For example, \xdc is exactly the same
5094 as \x{dc} (or \u00dc in JavaScript mode).
5096 Constraints on character values
5098 Characters that are specified using octal or hexadecimal numbers are
5099 limited to certain values, as follows:
5101 8-bit non-UTF mode less than 0x100
5102 8-bit UTF-8 mode less than 0x10ffff and a valid codepoint
5103 16-bit non-UTF mode less than 0x10000
5104 16-bit UTF-16 mode less than 0x10ffff and a valid codepoint
5105 32-bit non-UTF mode less than 0x100000000
5106 32-bit UTF-32 mode less than 0x10ffff and a valid codepoint
5108 Invalid Unicode codepoints are the range 0xd800 to 0xdfff (the so-
5109 called "surrogate" codepoints), and 0xffef.
5111 Escape sequences in character classes
5113 All the sequences that define a single character value can be used both
5114 inside and outside character classes. In addition, inside a character
5115 class, \b is interpreted as the backspace character (hex 08).
5117 \N is not allowed in a character class. \B, \R, and \X are not special
5118 inside a character class. Like other unrecognized escape sequences,
5119 they are treated as the literal characters "B", "R", and "X" by
5120 default, but cause an error if the PCRE_EXTRA option is set. Outside a
5121 character class, these sequences have different meanings.
5123 Unsupported escape sequences
5125 In Perl, the sequences \l, \L, \u, and \U are recognized by its string
5126 handler and used to modify the case of following characters. By
5127 default, PCRE does not support these escape sequences. However, if the
5128 PCRE_JAVASCRIPT_COMPAT option is set, \U matches a "U" character, and
5129 \u can be used to define a character by code point, as described in the
5132 Absolute and relative back references
5134 The sequence \g followed by an unsigned or a negative number, option-
5135 ally enclosed in braces, is an absolute or relative back reference. A
5136 named back reference can be coded as \g{name}. Back references are dis-
5137 cussed later, following the discussion of parenthesized subpatterns.
5139 Absolute and relative subroutine calls
5141 For compatibility with Oniguruma, the non-Perl syntax \g followed by a
5142 name or a number enclosed either in angle brackets or single quotes, is
5143 an alternative syntax for referencing a subpattern as a "subroutine".
5144 Details are discussed later. Note that \g{...} (Perl syntax) and
5145 \g<...> (Oniguruma syntax) are not synonymous. The former is a back
5146 reference; the latter is a subroutine call.
5148 Generic character types
5150 Another use of backslash is for specifying generic character types:
5152 \d any decimal digit
5153 \D any character that is not a decimal digit
5154 \h any horizontal white space character
5155 \H any character that is not a horizontal white space character
5156 \s any white space character
5157 \S any character that is not a white space character
5158 \v any vertical white space character
5159 \V any character that is not a vertical white space character
5160 \w any "word" character
5161 \W any "non-word" character
5163 There is also the single sequence \N, which matches a non-newline char-
5164 acter. This is the same as the "." metacharacter when PCRE_DOTALL is
5165 not set. Perl also uses \N to match characters by name; PCRE does not
5168 Each pair of lower and upper case escape sequences partitions the com-
5169 plete set of characters into two disjoint sets. Any given character
5170 matches one, and only one, of each pair. The sequences can appear both
5171 inside and outside character classes. They each match one character of
5172 the appropriate type. If the current matching point is at the end of
5173 the subject string, all of them fail, because there is no character to
5176 For compatibility with Perl, \s did not used to match the VT character
5177 (code 11), which made it different from the the POSIX "space" class.
5178 However, Perl added VT at release 5.18, and PCRE followed suit at
5179 release 8.34. The default \s characters are now HT (9), LF (10), VT
5180 (11), FF (12), CR (13), and space (32), which are defined as white
5181 space in the "C" locale. This list may vary if locale-specific matching
5182 is taking place. For example, in some locales the "non-breaking space"
5183 character (\xA0) is recognized as white space, and in others the VT
5186 A "word" character is an underscore or any character that is a letter
5187 or digit. By default, the definition of letters and digits is con-
5188 trolled by PCRE's low-valued character tables, and may vary if locale-
5189 specific matching is taking place (see "Locale support" in the pcreapi
5190 page). For example, in a French locale such as "fr_FR" in Unix-like
5191 systems, or "french" in Windows, some character codes greater than 127
5192 are used for accented letters, and these are then matched by \w. The
5193 use of locales with Unicode is discouraged.
5195 By default, characters whose code points are greater than 127 never
5196 match \d, \s, or \w, and always match \D, \S, and \W, although this may
5197 vary for characters in the range 128-255 when locale-specific matching
5198 is happening. These escape sequences retain their original meanings
5199 from before Unicode support was available, mainly for efficiency rea-
5200 sons. If PCRE is compiled with Unicode property support, and the
5201 PCRE_UCP option is set, the behaviour is changed so that Unicode prop-
5202 erties are used to determine character types, as follows:
5204 \d any character that matches \p{Nd} (decimal digit)
5205 \s any character that matches \p{Z} or \h or \v
5206 \w any character that matches \p{L} or \p{N}, plus underscore
5208 The upper case escapes match the inverse sets of characters. Note that
5209 \d matches only decimal digits, whereas \w matches any Unicode digit,
5210 as well as any Unicode letter, and underscore. Note also that PCRE_UCP
5211 affects \b, and \B because they are defined in terms of \w and \W.
5212 Matching these sequences is noticeably slower when PCRE_UCP is set.
5214 The sequences \h, \H, \v, and \V are features that were added to Perl
5215 at release 5.10. In contrast to the other sequences, which match only
5216 ASCII characters by default, these always match certain high-valued
5217 code points, whether or not PCRE_UCP is set. The horizontal space char-
5220 U+0009 Horizontal tab (HT)
5222 U+00A0 Non-break space
5223 U+1680 Ogham space mark
5224 U+180E Mongolian vowel separator
5229 U+2004 Three-per-em space
5230 U+2005 Four-per-em space
5231 U+2006 Six-per-em space
5233 U+2008 Punctuation space
5236 U+202F Narrow no-break space
5237 U+205F Medium mathematical space
5238 U+3000 Ideographic space
5240 The vertical space characters are:
5242 U+000A Linefeed (LF)
5243 U+000B Vertical tab (VT)
5244 U+000C Form feed (FF)
5245 U+000D Carriage return (CR)
5246 U+0085 Next line (NEL)
5247 U+2028 Line separator
5248 U+2029 Paragraph separator
5250 In 8-bit, non-UTF-8 mode, only the characters with codepoints less than
5255 Outside a character class, by default, the escape sequence \R matches
5256 any Unicode newline sequence. In 8-bit non-UTF-8 mode \R is equivalent
5259 (?>\r\n|\n|\x0b|\f|\r|\x85)
5261 This is an example of an "atomic group", details of which are given
5262 below. This particular group matches either the two-character sequence
5263 CR followed by LF, or one of the single characters LF (linefeed,
5264 U+000A), VT (vertical tab, U+000B), FF (form feed, U+000C), CR (car-
5265 riage return, U+000D), or NEL (next line, U+0085). The two-character
5266 sequence is treated as a single unit that cannot be split.
5268 In other modes, two additional characters whose codepoints are greater
5269 than 255 are added: LS (line separator, U+2028) and PS (paragraph sepa-
5270 rator, U+2029). Unicode character property support is not needed for
5271 these characters to be recognized.
5273 It is possible to restrict \R to match only CR, LF, or CRLF (instead of
5274 the complete set of Unicode line endings) by setting the option
5275 PCRE_BSR_ANYCRLF either at compile time or when the pattern is matched.
5276 (BSR is an abbrevation for "backslash R".) This can be made the default
5277 when PCRE is built; if this is the case, the other behaviour can be
5278 requested via the PCRE_BSR_UNICODE option. It is also possible to
5279 specify these settings by starting a pattern string with one of the
5280 following sequences:
5282 (*BSR_ANYCRLF) CR, LF, or CRLF only
5283 (*BSR_UNICODE) any Unicode newline sequence
5285 These override the default and the options given to the compiling func-
5286 tion, but they can themselves be overridden by options given to a
5287 matching function. Note that these special settings, which are not
5288 Perl-compatible, are recognized only at the very start of a pattern,
5289 and that they must be in upper case. If more than one of them is
5290 present, the last one is used. They can be combined with a change of
5291 newline convention; for example, a pattern can start with:
5293 (*ANY)(*BSR_ANYCRLF)
5295 They can also be combined with the (*UTF8), (*UTF16), (*UTF32), (*UTF)
5296 or (*UCP) special sequences. Inside a character class, \R is treated as
5297 an unrecognized escape sequence, and so matches the letter "R" by
5298 default, but causes an error if PCRE_EXTRA is set.
5300 Unicode character properties
5302 When PCRE is built with Unicode character property support, three addi-
5303 tional escape sequences that match characters with specific properties
5304 are available. When in 8-bit non-UTF-8 mode, these sequences are of
5305 course limited to testing characters whose codepoints are less than
5306 256, but they do work in this mode. The extra escape sequences are:
5308 \p{xx} a character with the xx property
5309 \P{xx} a character without the xx property
5310 \X a Unicode extended grapheme cluster
5312 The property names represented by xx above are limited to the Unicode
5313 script names, the general category properties, "Any", which matches any
5314 character (including newline), and some special PCRE properties
5315 (described in the next section). Other Perl properties such as "InMu-
5316 sicalSymbols" are not currently supported by PCRE. Note that \P{Any}
5317 does not match any characters, so always causes a match failure.
5319 Sets of Unicode characters are defined as belonging to certain scripts.
5320 A character from one of these sets can be matched using a script name.
5326 Those that are not part of an identified script are lumped together as
5327 "Common". The current list of scripts is:
5329 Arabic, Armenian, Avestan, Balinese, Bamum, Batak, Bengali, Bopomofo,
5330 Brahmi, Braille, Buginese, Buhid, Canadian_Aboriginal, Carian, Chakma,
5331 Cham, Cherokee, Common, Coptic, Cuneiform, Cypriot, Cyrillic, Deseret,
5332 Devanagari, Egyptian_Hieroglyphs, Ethiopic, Georgian, Glagolitic,
5333 Gothic, Greek, Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hira-
5334 gana, Imperial_Aramaic, Inherited, Inscriptional_Pahlavi, Inscrip-
5335 tional_Parthian, Javanese, Kaithi, Kannada, Katakana, Kayah_Li,
5336 Kharoshthi, Khmer, Lao, Latin, Lepcha, Limbu, Linear_B, Lisu, Lycian,
5337 Lydian, Malayalam, Mandaic, Meetei_Mayek, Meroitic_Cursive,
5338 Meroitic_Hieroglyphs, Miao, Mongolian, Myanmar, New_Tai_Lue, Nko,
5339 Ogham, Old_Italic, Old_Persian, Old_South_Arabian, Old_Turkic,
5340 Ol_Chiki, Oriya, Osmanya, Phags_Pa, Phoenician, Rejang, Runic, Samari-
5341 tan, Saurashtra, Sharada, Shavian, Sinhala, Sora_Sompeng, Sundanese,
5342 Syloti_Nagri, Syriac, Tagalog, Tagbanwa, Tai_Le, Tai_Tham, Tai_Viet,
5343 Takri, Tamil, Telugu, Thaana, Thai, Tibetan, Tifinagh, Ugaritic, Vai,
5346 Each character has exactly one Unicode general category property, spec-
5347 ified by a two-letter abbreviation. For compatibility with Perl, nega-
5348 tion can be specified by including a circumflex between the opening
5349 brace and the property name. For example, \p{^Lu} is the same as
5352 If only one letter is specified with \p or \P, it includes all the gen-
5353 eral category properties that start with that letter. In this case, in
5354 the absence of negation, the curly brackets in the escape sequence are
5355 optional; these two examples have the same effect:
5360 The following general category property codes are supported:
5370 Ll Lower case letter
5373 Lt Title case letter
5374 Lu Upper case letter
5387 Pc Connector punctuation
5389 Pe Close punctuation
5390 Pf Final punctuation
5391 Pi Initial punctuation
5392 Po Other punctuation
5398 Sm Mathematical symbol
5403 Zp Paragraph separator
5406 The special property L& is also supported: it matches a character that
5407 has the Lu, Ll, or Lt property, in other words, a letter that is not
5408 classified as a modifier or "other".
5410 The Cs (Surrogate) property applies only to characters in the range
5411 U+D800 to U+DFFF. Such characters are not valid in Unicode strings and
5412 so cannot be tested by PCRE, unless UTF validity checking has been
5413 turned off (see the discussion of PCRE_NO_UTF8_CHECK,
5414 PCRE_NO_UTF16_CHECK and PCRE_NO_UTF32_CHECK in the pcreapi page). Perl
5415 does not support the Cs property.
5417 The long synonyms for property names that Perl supports (such as
5418 \p{Letter}) are not supported by PCRE, nor is it permitted to prefix
5419 any of these properties with "Is".
5421 No character that is in the Unicode table has the Cn (unassigned) prop-
5422 erty. Instead, this property is assumed for any code point that is not
5423 in the Unicode table.
5425 Specifying caseless matching does not affect these escape sequences.
5426 For example, \p{Lu} always matches only upper case letters. This is
5427 different from the behaviour of current versions of Perl.
5429 Matching characters by Unicode property is not fast, because PCRE has
5430 to do a multistage table lookup in order to find a character's prop-
5431 erty. That is why the traditional escape sequences such as \d and \w do
5432 not use Unicode properties in PCRE by default, though you can make them
5433 do so by setting the PCRE_UCP option or by starting the pattern with
5436 Extended grapheme clusters
5438 The \X escape matches any number of Unicode characters that form an
5439 "extended grapheme cluster", and treats the sequence as an atomic group
5440 (see below). Up to and including release 8.31, PCRE matched an ear-
5441 lier, simpler definition that was equivalent to
5445 That is, it matched a character without the "mark" property, followed
5446 by zero or more characters with the "mark" property. Characters with
5447 the "mark" property are typically non-spacing accents that affect the
5448 preceding character.
5450 This simple definition was extended in Unicode to include more compli-
5451 cated kinds of composite character by giving each character a grapheme
5452 breaking property, and creating rules that use these properties to
5453 define the boundaries of extended grapheme clusters. In releases of
5454 PCRE later than 8.31, \X matches one of these clusters.
5456 \X always matches at least one character. Then it decides whether to
5457 add additional characters according to the following rules for ending a
5460 1. End at the end of the subject string.
5462 2. Do not end between CR and LF; otherwise end after any control char-
5465 3. Do not break Hangul (a Korean script) syllable sequences. Hangul
5466 characters are of five types: L, V, T, LV, and LVT. An L character may
5467 be followed by an L, V, LV, or LVT character; an LV or V character may
5468 be followed by a V or T character; an LVT or T character may be follwed
5469 only by a T character.
5471 4. Do not end before extending characters or spacing marks. Characters
5472 with the "mark" property always have the "extend" grapheme breaking
5475 5. Do not end after prepend characters.
5477 6. Otherwise, end the cluster.
5479 PCRE's additional properties
5481 As well as the standard Unicode properties described above, PCRE sup-
5482 ports four more that make it possible to convert traditional escape
5483 sequences such as \w and \s to use Unicode properties. PCRE uses these
5484 non-standard, non-Perl properties internally when PCRE_UCP is set. How-
5485 ever, they may also be used explicitly. These properties are:
5487 Xan Any alphanumeric character
5488 Xps Any POSIX space character
5489 Xsp Any Perl space character
5490 Xwd Any Perl "word" character
5492 Xan matches characters that have either the L (letter) or the N (num-
5493 ber) property. Xps matches the characters tab, linefeed, vertical tab,
5494 form feed, or carriage return, and any other character that has the Z
5495 (separator) property. Xsp is the same as Xps; it used to exclude ver-
5496 tical tab, for Perl compatibility, but Perl changed, and so PCRE fol-
5497 lowed at release 8.34. Xwd matches the same characters as Xan, plus
5500 There is another non-standard property, Xuc, which matches any charac-
5501 ter that can be represented by a Universal Character Name in C++ and
5502 other programming languages. These are the characters $, @, ` (grave
5503 accent), and all characters with Unicode code points greater than or
5504 equal to U+00A0, except for the surrogates U+D800 to U+DFFF. Note that
5505 most base (ASCII) characters are excluded. (Universal Character Names
5506 are of the form \uHHHH or \UHHHHHHHH where H is a hexadecimal digit.
5507 Note that the Xuc property does not match these sequences but the char-
5508 acters that they represent.)
5510 Resetting the match start
5512 The escape sequence \K causes any previously matched characters not to
5513 be included in the final matched sequence. For example, the pattern:
5517 matches "foobar", but reports that it has matched "bar". This feature
5518 is similar to a lookbehind assertion (described below). However, in
5519 this case, the part of the subject before the real match does not have
5520 to be of fixed length, as lookbehind assertions do. The use of \K does
5521 not interfere with the setting of captured substrings. For example,
5526 matches "foobar", the first substring is still set to "foo".
5528 Perl documents that the use of \K within assertions is "not well
5529 defined". In PCRE, \K is acted upon when it occurs inside positive
5530 assertions, but is ignored in negative assertions. Note that when a
5531 pattern such as (?=ab\K) matches, the reported start of the match can
5532 be greater than the end of the match.
5536 The final use of backslash is for certain simple assertions. An asser-
5537 tion specifies a condition that has to be met at a particular point in
5538 a match, without consuming any characters from the subject string. The
5539 use of subpatterns for more complicated assertions is described below.
5540 The backslashed assertions are:
5542 \b matches at a word boundary
5543 \B matches when not at a word boundary
5544 \A matches at the start of the subject
5545 \Z matches at the end of the subject
5546 also matches before a newline at the end of the subject
5547 \z matches only at the end of the subject
5548 \G matches at the first matching position in the subject
5550 Inside a character class, \b has a different meaning; it matches the
5551 backspace character. If any other of these assertions appears in a
5552 character class, by default it matches the corresponding literal char-
5553 acter (for example, \B matches the letter B). However, if the
5554 PCRE_EXTRA option is set, an "invalid escape sequence" error is gener-
5557 A word boundary is a position in the subject string where the current
5558 character and the previous character do not both match \w or \W (i.e.
5559 one matches \w and the other matches \W), or the start or end of the
5560 string if the first or last character matches \w, respectively. In a
5561 UTF mode, the meanings of \w and \W can be changed by setting the
5562 PCRE_UCP option. When this is done, it also affects \b and \B. Neither
5563 PCRE nor Perl has a separate "start of word" or "end of word" metase-
5564 quence. However, whatever follows \b normally determines which it is.
5565 For example, the fragment \ba matches "a" at the start of a word.
5567 The \A, \Z, and \z assertions differ from the traditional circumflex
5568 and dollar (described in the next section) in that they only ever match
5569 at the very start and end of the subject string, whatever options are
5570 set. Thus, they are independent of multiline mode. These three asser-
5571 tions are not affected by the PCRE_NOTBOL or PCRE_NOTEOL options, which
5572 affect only the behaviour of the circumflex and dollar metacharacters.
5573 However, if the startoffset argument of pcre_exec() is non-zero, indi-
5574 cating that matching is to start at a point other than the beginning of
5575 the subject, \A can never match. The difference between \Z and \z is
5576 that \Z matches before a newline at the end of the string as well as at
5577 the very end, whereas \z matches only at the end.
5579 The \G assertion is true only when the current matching position is at
5580 the start point of the match, as specified by the startoffset argument
5581 of pcre_exec(). It differs from \A when the value of startoffset is
5582 non-zero. By calling pcre_exec() multiple times with appropriate argu-
5583 ments, you can mimic Perl's /g option, and it is in this kind of imple-
5584 mentation where \G can be useful.
5586 Note, however, that PCRE's interpretation of \G, as the start of the
5587 current match, is subtly different from Perl's, which defines it as the
5588 end of the previous match. In Perl, these can be different when the
5589 previously matched string was empty. Because PCRE does just one match
5590 at a time, it cannot reproduce this behaviour.
5592 If all the alternatives of a pattern begin with \G, the expression is
5593 anchored to the starting match position, and the "anchored" flag is set
5594 in the compiled regular expression.
5597 CIRCUMFLEX AND DOLLAR
5599 The circumflex and dollar metacharacters are zero-width assertions.
5600 That is, they test for a particular condition being true without con-
5601 suming any characters from the subject string.
5603 Outside a character class, in the default matching mode, the circumflex
5604 character is an assertion that is true only if the current matching
5605 point is at the start of the subject string. If the startoffset argu-
5606 ment of pcre_exec() is non-zero, circumflex can never match if the
5607 PCRE_MULTILINE option is unset. Inside a character class, circumflex
5608 has an entirely different meaning (see below).
5610 Circumflex need not be the first character of the pattern if a number
5611 of alternatives are involved, but it should be the first thing in each
5612 alternative in which it appears if the pattern is ever to match that
5613 branch. If all possible alternatives start with a circumflex, that is,
5614 if the pattern is constrained to match only at the start of the sub-
5615 ject, it is said to be an "anchored" pattern. (There are also other
5616 constructs that can cause a pattern to be anchored.)
5618 The dollar character is an assertion that is true only if the current
5619 matching point is at the end of the subject string, or immediately
5620 before a newline at the end of the string (by default). Note, however,
5621 that it does not actually match the newline. Dollar need not be the
5622 last character of the pattern if a number of alternatives are involved,
5623 but it should be the last item in any branch in which it appears. Dol-
5624 lar has no special meaning in a character class.
5626 The meaning of dollar can be changed so that it matches only at the
5627 very end of the string, by setting the PCRE_DOLLAR_ENDONLY option at
5628 compile time. This does not affect the \Z assertion.
5630 The meanings of the circumflex and dollar characters are changed if the
5631 PCRE_MULTILINE option is set. When this is the case, a circumflex
5632 matches immediately after internal newlines as well as at the start of
5633 the subject string. It does not match after a newline that ends the
5634 string. A dollar matches before any newlines in the string, as well as
5635 at the very end, when PCRE_MULTILINE is set. When newline is specified
5636 as the two-character sequence CRLF, isolated CR and LF characters do
5637 not indicate newlines.
5639 For example, the pattern /^abc$/ matches the subject string "def\nabc"
5640 (where \n represents a newline) in multiline mode, but not otherwise.
5641 Consequently, patterns that are anchored in single line mode because
5642 all branches start with ^ are not anchored in multiline mode, and a
5643 match for circumflex is possible when the startoffset argument of
5644 pcre_exec() is non-zero. The PCRE_DOLLAR_ENDONLY option is ignored if
5645 PCRE_MULTILINE is set.
5647 Note that the sequences \A, \Z, and \z can be used to match the start
5648 and end of the subject in both modes, and if all branches of a pattern
5649 start with \A it is always anchored, whether or not PCRE_MULTILINE is
5653 FULL STOP (PERIOD, DOT) AND \N
5655 Outside a character class, a dot in the pattern matches any one charac-
5656 ter in the subject string except (by default) a character that signi-
5657 fies the end of a line.
5659 When a line ending is defined as a single character, dot never matches
5660 that character; when the two-character sequence CRLF is used, dot does
5661 not match CR if it is immediately followed by LF, but otherwise it
5662 matches all characters (including isolated CRs and LFs). When any Uni-
5663 code line endings are being recognized, dot does not match CR or LF or
5664 any of the other line ending characters.
5666 The behaviour of dot with regard to newlines can be changed. If the
5667 PCRE_DOTALL option is set, a dot matches any one character, without
5668 exception. If the two-character sequence CRLF is present in the subject
5669 string, it takes two dots to match it.
5671 The handling of dot is entirely independent of the handling of circum-
5672 flex and dollar, the only relationship being that they both involve
5673 newlines. Dot has no special meaning in a character class.
5675 The escape sequence \N behaves like a dot, except that it is not
5676 affected by the PCRE_DOTALL option. In other words, it matches any
5677 character except one that signifies the end of a line. Perl also uses
5678 \N to match characters by name; PCRE does not support this.
5681 MATCHING A SINGLE DATA UNIT
5683 Outside a character class, the escape sequence \C matches any one data
5684 unit, whether or not a UTF mode is set. In the 8-bit library, one data
5685 unit is one byte; in the 16-bit library it is a 16-bit unit; in the
5686 32-bit library it is a 32-bit unit. Unlike a dot, \C always matches
5687 line-ending characters. The feature is provided in Perl in order to
5688 match individual bytes in UTF-8 mode, but it is unclear how it can use-
5689 fully be used. Because \C breaks up characters into individual data
5690 units, matching one unit with \C in a UTF mode means that the rest of
5691 the string may start with a malformed UTF character. This has undefined
5692 results, because PCRE assumes that it is dealing with valid UTF strings
5693 (and by default it checks this at the start of processing unless the
5694 PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK or PCRE_NO_UTF32_CHECK option
5697 PCRE does not allow \C to appear in lookbehind assertions (described
5698 below) in a UTF mode, because this would make it impossible to calcu-
5699 late the length of the lookbehind.
5701 In general, the \C escape sequence is best avoided. However, one way of
5702 using it that avoids the problem of malformed UTF characters is to use
5703 a lookahead to check the length of the next character, as in this pat-
5704 tern, which could be used with a UTF-8 string (ignore white space and
5707 (?| (?=[\x00-\x7f])(\C) |
5708 (?=[\x80-\x{7ff}])(\C)(\C) |
5709 (?=[\x{800}-\x{ffff}])(\C)(\C)(\C) |
5710 (?=[\x{10000}-\x{1fffff}])(\C)(\C)(\C)(\C))
5712 A group that starts with (?| resets the capturing parentheses numbers
5713 in each alternative (see "Duplicate Subpattern Numbers" below). The
5714 assertions at the start of each branch check the next UTF-8 character
5715 for values whose encoding uses 1, 2, 3, or 4 bytes, respectively. The
5716 character's individual bytes are then captured by the appropriate num-
5720 SQUARE BRACKETS AND CHARACTER CLASSES
5722 An opening square bracket introduces a character class, terminated by a
5723 closing square bracket. A closing square bracket on its own is not spe-
5724 cial by default. However, if the PCRE_JAVASCRIPT_COMPAT option is set,
5725 a lone closing square bracket causes a compile-time error. If a closing
5726 square bracket is required as a member of the class, it should be the
5727 first data character in the class (after an initial circumflex, if
5728 present) or escaped with a backslash.
5730 A character class matches a single character in the subject. In a UTF
5731 mode, the character may be more than one data unit long. A matched
5732 character must be in the set of characters defined by the class, unless
5733 the first character in the class definition is a circumflex, in which
5734 case the subject character must not be in the set defined by the class.
5735 If a circumflex is actually required as a member of the class, ensure
5736 it is not the first character, or escape it with a backslash.
5738 For example, the character class [aeiou] matches any lower case vowel,
5739 while [^aeiou] matches any character that is not a lower case vowel.
5740 Note that a circumflex is just a convenient notation for specifying the
5741 characters that are in the class by enumerating those that are not. A
5742 class that starts with a circumflex is not an assertion; it still con-
5743 sumes a character from the subject string, and therefore it fails if
5744 the current pointer is at the end of the string.
5746 In UTF-8 (UTF-16, UTF-32) mode, characters with values greater than 255
5747 (0xffff) can be included in a class as a literal string of data units,
5748 or by using the \x{ escaping mechanism.
5750 When caseless matching is set, any letters in a class represent both
5751 their upper case and lower case versions, so for example, a caseless
5752 [aeiou] matches "A" as well as "a", and a caseless [^aeiou] does not
5753 match "A", whereas a caseful version would. In a UTF mode, PCRE always
5754 understands the concept of case for characters whose values are less
5755 than 128, so caseless matching is always possible. For characters with
5756 higher values, the concept of case is supported if PCRE is compiled
5757 with Unicode property support, but not otherwise. If you want to use
5758 caseless matching in a UTF mode for characters 128 and above, you must
5759 ensure that PCRE is compiled with Unicode property support as well as
5762 Characters that might indicate line breaks are never treated in any
5763 special way when matching character classes, whatever line-ending
5764 sequence is in use, and whatever setting of the PCRE_DOTALL and
5765 PCRE_MULTILINE options is used. A class such as [^a] always matches one
5766 of these characters.
5768 The minus (hyphen) character can be used to specify a range of charac-
5769 ters in a character class. For example, [d-m] matches any letter
5770 between d and m, inclusive. If a minus character is required in a
5771 class, it must be escaped with a backslash or appear in a position
5772 where it cannot be interpreted as indicating a range, typically as the
5773 first or last character in the class, or immediately after a range. For
5774 example, [b-d-z] matches letters in the range b to d, a hyphen charac-
5777 It is not possible to have the literal character "]" as the end charac-
5778 ter of a range. A pattern such as [W-]46] is interpreted as a class of
5779 two characters ("W" and "-") followed by a literal string "46]", so it
5780 would match "W46]" or "-46]". However, if the "]" is escaped with a
5781 backslash it is interpreted as the end of range, so [W-\]46] is inter-
5782 preted as a class containing a range followed by two other characters.
5783 The octal or hexadecimal representation of "]" can also be used to end
5786 An error is generated if a POSIX character class (see below) or an
5787 escape sequence other than one that defines a single character appears
5788 at a point where a range ending character is expected. For example,
5789 [z-\xff] is valid, but [A-\d] and [A-[:digit:]] are not.
5791 Ranges operate in the collating sequence of character values. They can
5792 also be used for characters specified numerically, for example
5793 [\000-\037]. Ranges can include any characters that are valid for the
5796 If a range that includes letters is used when caseless matching is set,
5797 it matches the letters in either case. For example, [W-c] is equivalent
5798 to [][\\^_`wxyzabc], matched caselessly, and in a non-UTF mode, if
5799 character tables for a French locale are in use, [\xc8-\xcb] matches
5800 accented E characters in both cases. In UTF modes, PCRE supports the
5801 concept of case for characters with values greater than 128 only when
5802 it is compiled with Unicode property support.
5804 The character escape sequences \d, \D, \h, \H, \p, \P, \s, \S, \v, \V,
5805 \w, and \W may appear in a character class, and add the characters that
5806 they match to the class. For example, [\dABCDEF] matches any hexadeci-
5807 mal digit. In UTF modes, the PCRE_UCP option affects the meanings of
5808 \d, \s, \w and their upper case partners, just as it does when they
5809 appear outside a character class, as described in the section entitled
5810 "Generic character types" above. The escape sequence \b has a different
5811 meaning inside a character class; it matches the backspace character.
5812 The sequences \B, \N, \R, and \X are not special inside a character
5813 class. Like any other unrecognized escape sequences, they are treated
5814 as the literal characters "B", "N", "R", and "X" by default, but cause
5815 an error if the PCRE_EXTRA option is set.
5817 A circumflex can conveniently be used with the upper case character
5818 types to specify a more restricted set of characters than the matching
5819 lower case type. For example, the class [^\W_] matches any letter or
5820 digit, but not underscore, whereas [\w] includes underscore. A positive
5821 character class should be read as "something OR something OR ..." and a
5822 negative class as "NOT something AND NOT something AND NOT ...".
5824 The only metacharacters that are recognized in character classes are
5825 backslash, hyphen (only where it can be interpreted as specifying a
5826 range), circumflex (only at the start), opening square bracket (only
5827 when it can be interpreted as introducing a POSIX class name, or for a
5828 special compatibility feature - see the next two sections), and the
5829 terminating closing square bracket. However, escaping other non-
5830 alphanumeric characters does no harm.
5833 POSIX CHARACTER CLASSES
5835 Perl supports the POSIX notation for character classes. This uses names
5836 enclosed by [: and :] within the enclosing square brackets. PCRE also
5837 supports this notation. For example,
5841 matches "0", "1", any alphabetic character, or "%". The supported class
5844 alnum letters and digits
5846 ascii character codes 0 - 127
5847 blank space or tab only
5848 cntrl control characters
5849 digit decimal digits (same as \d)
5850 graph printing characters, excluding space
5851 lower lower case letters
5852 print printing characters, including space
5853 punct printing characters, excluding letters and digits and space
5854 space white space (the same as \s from PCRE 8.34)
5855 upper upper case letters
5856 word "word" characters (same as \w)
5857 xdigit hexadecimal digits
5859 The default "space" characters are HT (9), LF (10), VT (11), FF (12),
5860 CR (13), and space (32). If locale-specific matching is taking place,
5861 the list of space characters may be different; there may be fewer or
5862 more of them. "Space" used to be different to \s, which did not include
5863 VT, for Perl compatibility. However, Perl changed at release 5.18, and
5864 PCRE followed at release 8.34. "Space" and \s now match the same set
5867 The name "word" is a Perl extension, and "blank" is a GNU extension
5868 from Perl 5.8. Another Perl extension is negation, which is indicated
5869 by a ^ character after the colon. For example,
5873 matches "1", "2", or any non-digit. PCRE (and Perl) also recognize the
5874 POSIX syntax [.ch.] and [=ch=] where "ch" is a "collating element", but
5875 these are not supported, and an error is given if they are encountered.
5877 By default, characters with values greater than 128 do not match any of
5878 the POSIX character classes. However, if the PCRE_UCP option is passed
5879 to pcre_compile(), some of the classes are changed so that Unicode
5880 character properties are used. This is achieved by replacing certain
5881 POSIX classes by other sequences, as follows:
5883 [:alnum:] becomes \p{Xan}
5884 [:alpha:] becomes \p{L}
5885 [:blank:] becomes \h
5886 [:digit:] becomes \p{Nd}
5887 [:lower:] becomes \p{Ll}
5888 [:space:] becomes \p{Xps}
5889 [:upper:] becomes \p{Lu}
5890 [:word:] becomes \p{Xwd}
5892 Negated versions, such as [:^alpha:] use \P instead of \p. Three other
5893 POSIX classes are handled specially in UCP mode:
5895 [:graph:] This matches characters that have glyphs that mark the page
5896 when printed. In Unicode property terms, it matches all char-
5897 acters with the L, M, N, P, S, or Cf properties, except for:
5899 U+061C Arabic Letter Mark
5900 U+180E Mongolian Vowel Separator
5901 U+2066 - U+2069 Various "isolate"s
5904 [:print:] This matches the same characters as [:graph:] plus space
5905 characters that are not controls, that is, characters with
5908 [:punct:] This matches all characters that have the Unicode P (punctua-
5909 tion) property, plus those characters whose code points are
5910 less than 128 that have the S (Symbol) property.
5912 The other POSIX classes are unchanged, and match only characters with
5913 code points less than 128.
5916 COMPATIBILITY FEATURE FOR WORD BOUNDARIES
5918 In the POSIX.2 compliant library that was included in 4.4BSD Unix, the
5919 ugly syntax [[:<:]] and [[:>:]] is used for matching "start of word"
5920 and "end of word". PCRE treats these items as follows:
5922 [[:<:]] is converted to \b(?=\w)
5923 [[:>:]] is converted to \b(?<=\w)
5925 Only these exact character sequences are recognized. A sequence such as
5926 [a[:<:]b] provokes error for an unrecognized POSIX class name. This
5927 support is not compatible with Perl. It is provided to help migrations
5928 from other environments, and is best not used in any new patterns. Note
5929 that \b matches at the start and the end of a word (see "Simple asser-
5930 tions" above), and in a Perl-style pattern the preceding or following
5931 character normally shows which is wanted, without the need for the
5932 assertions that are used above in order to give exactly the POSIX be-
5938 Vertical bar characters are used to separate alternative patterns. For
5939 example, the pattern
5943 matches either "gilbert" or "sullivan". Any number of alternatives may
5944 appear, and an empty alternative is permitted (matching the empty
5945 string). The matching process tries each alternative in turn, from left
5946 to right, and the first one that succeeds is used. If the alternatives
5947 are within a subpattern (defined below), "succeeds" means matching the
5948 rest of the main pattern as well as the alternative in the subpattern.
5951 INTERNAL OPTION SETTING
5953 The settings of the PCRE_CASELESS, PCRE_MULTILINE, PCRE_DOTALL, and
5954 PCRE_EXTENDED options (which are Perl-compatible) can be changed from
5955 within the pattern by a sequence of Perl option letters enclosed
5956 between "(?" and ")". The option letters are
5959 m for PCRE_MULTILINE
5963 For example, (?im) sets caseless, multiline matching. It is also possi-
5964 ble to unset these options by preceding the letter with a hyphen, and a
5965 combined setting and unsetting such as (?im-sx), which sets PCRE_CASE-
5966 LESS and PCRE_MULTILINE while unsetting PCRE_DOTALL and PCRE_EXTENDED,
5967 is also permitted. If a letter appears both before and after the
5968 hyphen, the option is unset.
5970 The PCRE-specific options PCRE_DUPNAMES, PCRE_UNGREEDY, and PCRE_EXTRA
5971 can be changed in the same way as the Perl-compatible options by using
5972 the characters J, U and X respectively.
5974 When one of these option changes occurs at top level (that is, not
5975 inside subpattern parentheses), the change applies to the remainder of
5976 the pattern that follows. If the change is placed right at the start of
5977 a pattern, PCRE extracts it into the global options (and it will there-
5978 fore show up in data extracted by the pcre_fullinfo() function).
5980 An option change within a subpattern (see below for a description of
5981 subpatterns) affects only that part of the subpattern that follows it,
5986 matches abc and aBc and no other strings (assuming PCRE_CASELESS is not
5987 used). By this means, options can be made to have different settings
5988 in different parts of the pattern. Any changes made in one alternative
5989 do carry on into subsequent branches within the same subpattern. For
5994 matches "ab", "aB", "c", and "C", even though when matching "C" the
5995 first branch is abandoned before the option setting. This is because
5996 the effects of option settings happen at compile time. There would be
5997 some very weird behaviour otherwise.
5999 Note: There are other PCRE-specific options that can be set by the
6000 application when the compiling or matching functions are called. In
6001 some cases the pattern can contain special leading sequences such as
6002 (*CRLF) to override what the application has set or what has been
6003 defaulted. Details are given in the section entitled "Newline
6004 sequences" above. There are also the (*UTF8), (*UTF16),(*UTF32), and
6005 (*UCP) leading sequences that can be used to set UTF and Unicode prop-
6006 erty modes; they are equivalent to setting the PCRE_UTF8, PCRE_UTF16,
6007 PCRE_UTF32 and the PCRE_UCP options, respectively. The (*UTF) sequence
6008 is a generic version that can be used with any of the libraries. How-
6009 ever, the application can set the PCRE_NEVER_UTF option, which locks
6010 out the use of the (*UTF) sequences.
6015 Subpatterns are delimited by parentheses (round brackets), which can be
6016 nested. Turning part of a pattern into a subpattern does two things:
6018 1. It localizes a set of alternatives. For example, the pattern
6020 cat(aract|erpillar|)
6022 matches "cataract", "caterpillar", or "cat". Without the parentheses,
6023 it would match "cataract", "erpillar" or an empty string.
6025 2. It sets up the subpattern as a capturing subpattern. This means
6026 that, when the whole pattern matches, that portion of the subject
6027 string that matched the subpattern is passed back to the caller via the
6028 ovector argument of the matching function. (This applies only to the
6029 traditional matching functions; the DFA matching functions do not sup-
6032 Opening parentheses are counted from left to right (starting from 1) to
6033 obtain numbers for the capturing subpatterns. For example, if the
6034 string "the red king" is matched against the pattern
6036 the ((red|white) (king|queen))
6038 the captured substrings are "red king", "red", and "king", and are num-
6039 bered 1, 2, and 3, respectively.
6041 The fact that plain parentheses fulfil two functions is not always
6042 helpful. There are often times when a grouping subpattern is required
6043 without a capturing requirement. If an opening parenthesis is followed
6044 by a question mark and a colon, the subpattern does not do any captur-
6045 ing, and is not counted when computing the number of any subsequent
6046 capturing subpatterns. For example, if the string "the white queen" is
6047 matched against the pattern
6049 the ((?:red|white) (king|queen))
6051 the captured substrings are "white queen" and "queen", and are numbered
6052 1 and 2. The maximum number of capturing subpatterns is 65535.
6054 As a convenient shorthand, if any option settings are required at the
6055 start of a non-capturing subpattern, the option letters may appear
6056 between the "?" and the ":". Thus the two patterns
6058 (?i:saturday|sunday)
6059 (?:(?i)saturday|sunday)
6061 match exactly the same set of strings. Because alternative branches are
6062 tried from left to right, and options are not reset until the end of
6063 the subpattern is reached, an option setting in one branch does affect
6064 subsequent branches, so the above patterns match "SUNDAY" as well as
6068 DUPLICATE SUBPATTERN NUMBERS
6070 Perl 5.10 introduced a feature whereby each alternative in a subpattern
6071 uses the same numbers for its capturing parentheses. Such a subpattern
6072 starts with (?| and is itself a non-capturing subpattern. For example,
6073 consider this pattern:
6075 (?|(Sat)ur|(Sun))day
6077 Because the two alternatives are inside a (?| group, both sets of cap-
6078 turing parentheses are numbered one. Thus, when the pattern matches,
6079 you can look at captured substring number one, whichever alternative
6080 matched. This construct is useful when you want to capture part, but
6081 not all, of one of a number of alternatives. Inside a (?| group, paren-
6082 theses are numbered as usual, but the number is reset at the start of
6083 each branch. The numbers of any capturing parentheses that follow the
6084 subpattern start after the highest number used in any branch. The fol-
6085 lowing example is taken from the Perl documentation. The numbers under-
6086 neath show in which buffer the captured content will be stored.
6088 # before ---------------branch-reset----------- after
6089 / ( a ) (?| x ( y ) z | (p (q) r) | (t) u (v) ) ( z ) /x
6092 A back reference to a numbered subpattern uses the most recent value
6093 that is set for that number by any subpattern. The following pattern
6094 matches "abcabc" or "defdef":
6098 In contrast, a subroutine call to a numbered subpattern always refers
6099 to the first one in the pattern with the given number. The following
6100 pattern matches "abcabc" or "defabc":
6102 /(?|(abc)|(def))(?1)/
6104 If a condition test for a subpattern's having matched refers to a non-
6105 unique number, the test is true if any of the subpatterns of that num-
6108 An alternative approach to using this "branch reset" feature is to use
6109 duplicate named subpatterns, as described in the next section.
6114 Identifying capturing parentheses by number is simple, but it can be
6115 very hard to keep track of the numbers in complicated regular expres-
6116 sions. Furthermore, if an expression is modified, the numbers may
6117 change. To help with this difficulty, PCRE supports the naming of sub-
6118 patterns. This feature was not added to Perl until release 5.10. Python
6119 had the feature earlier, and PCRE introduced it at release 4.0, using
6120 the Python syntax. PCRE now supports both the Perl and the Python syn-
6121 tax. Perl allows identically numbered subpatterns to have different
6122 names, but PCRE does not.
6124 In PCRE, a subpattern can be named in one of three ways: (?<name>...)
6125 or (?'name'...) as in Perl, or (?P<name>...) as in Python. References
6126 to capturing parentheses from other parts of the pattern, such as back
6127 references, recursion, and conditions, can be made by name as well as
6130 Names consist of up to 32 alphanumeric characters and underscores, but
6131 must start with a non-digit. Named capturing parentheses are still
6132 allocated numbers as well as names, exactly as if the names were not
6133 present. The PCRE API provides function calls for extracting the name-
6134 to-number translation table from a compiled pattern. There is also a
6135 convenience function for extracting a captured substring by name.
6137 By default, a name must be unique within a pattern, but it is possible
6138 to relax this constraint by setting the PCRE_DUPNAMES option at compile
6139 time. (Duplicate names are also always permitted for subpatterns with
6140 the same number, set up as described in the previous section.) Dupli-
6141 cate names can be useful for patterns where only one instance of the
6142 named parentheses can match. Suppose you want to match the name of a
6143 weekday, either as a 3-letter abbreviation or as the full name, and in
6144 both cases you want to extract the abbreviation. This pattern (ignoring
6145 the line breaks) does the job:
6147 (?<DN>Mon|Fri|Sun)(?:day)?|
6148 (?<DN>Tue)(?:sday)?|
6149 (?<DN>Wed)(?:nesday)?|
6150 (?<DN>Thu)(?:rsday)?|
6151 (?<DN>Sat)(?:urday)?
6153 There are five capturing substrings, but only one is ever set after a
6154 match. (An alternative way of solving this problem is to use a "branch
6155 reset" subpattern, as described in the previous section.)
6157 The convenience function for extracting the data by name returns the
6158 substring for the first (and in this example, the only) subpattern of
6159 that name that matched. This saves searching to find which numbered
6162 If you make a back reference to a non-unique named subpattern from
6163 elsewhere in the pattern, the subpatterns to which the name refers are
6164 checked in the order in which they appear in the overall pattern. The
6165 first one that is set is used for the reference. For example, this pat-
6166 tern matches both "foofoo" and "barbar" but not "foobar" or "barfoo":
6168 (?:(?<n>foo)|(?<n>bar))\k<n>
6171 If you make a subroutine call to a non-unique named subpattern, the one
6172 that corresponds to the first occurrence of the name is used. In the
6173 absence of duplicate numbers (see the previous section) this is the one
6174 with the lowest number.
6176 If you use a named reference in a condition test (see the section about
6177 conditions below), either to check whether a subpattern has matched, or
6178 to check for recursion, all subpatterns with the same name are tested.
6179 If the condition is true for any one of them, the overall condition is
6180 true. This is the same behaviour as testing by number. For further
6181 details of the interfaces for handling named subpatterns, see the
6182 pcreapi documentation.
6184 Warning: You cannot use different names to distinguish between two sub-
6185 patterns with the same number because PCRE uses only the numbers when
6186 matching. For this reason, an error is given at compile time if differ-
6187 ent names are given to subpatterns with the same number. However, you
6188 can always give the same name to subpatterns with the same number, even
6189 when PCRE_DUPNAMES is not set.
6194 Repetition is specified by quantifiers, which can follow any of the
6197 a literal data character
6198 the dot metacharacter
6199 the \C escape sequence
6200 the \X escape sequence
6201 the \R escape sequence
6202 an escape such as \d or \pL that matches a single character
6204 a back reference (see next section)
6205 a parenthesized subpattern (including assertions)
6206 a subroutine call to a subpattern (recursive or otherwise)
6208 The general repetition quantifier specifies a minimum and maximum num-
6209 ber of permitted matches, by giving the two numbers in curly brackets
6210 (braces), separated by a comma. The numbers must be less than 65536,
6211 and the first must be less than or equal to the second. For example:
6215 matches "zz", "zzz", or "zzzz". A closing brace on its own is not a
6216 special character. If the second number is omitted, but the comma is
6217 present, there is no upper limit; if the second number and the comma
6218 are both omitted, the quantifier specifies an exact number of required
6223 matches at least 3 successive vowels, but may match many more, while
6227 matches exactly 8 digits. An opening curly bracket that appears in a
6228 position where a quantifier is not allowed, or one that does not match
6229 the syntax of a quantifier, is taken as a literal character. For exam-
6230 ple, {,6} is not a quantifier, but a literal string of four characters.
6232 In UTF modes, quantifiers apply to characters rather than to individual
6233 data units. Thus, for example, \x{100}{2} matches two characters, each
6234 of which is represented by a two-byte sequence in a UTF-8 string. Simi-
6235 larly, \X{3} matches three Unicode extended grapheme clusters, each of
6236 which may be several data units long (and they may be of different
6239 The quantifier {0} is permitted, causing the expression to behave as if
6240 the previous item and the quantifier were not present. This may be use-
6241 ful for subpatterns that are referenced as subroutines from elsewhere
6242 in the pattern (but see also the section entitled "Defining subpatterns
6243 for use by reference only" below). Items other than subpatterns that
6244 have a {0} quantifier are omitted from the compiled pattern.
6246 For convenience, the three most common quantifiers have single-charac-
6249 * is equivalent to {0,}
6250 + is equivalent to {1,}
6251 ? is equivalent to {0,1}
6253 It is possible to construct infinite loops by following a subpattern
6254 that can match no characters with a quantifier that has no upper limit,
6259 Earlier versions of Perl and PCRE used to give an error at compile time
6260 for such patterns. However, because there are cases where this can be
6261 useful, such patterns are now accepted, but if any repetition of the
6262 subpattern does in fact match no characters, the loop is forcibly bro-
6265 By default, the quantifiers are "greedy", that is, they match as much
6266 as possible (up to the maximum number of permitted times), without
6267 causing the rest of the pattern to fail. The classic example of where
6268 this gives problems is in trying to match comments in C programs. These
6269 appear between /* and */ and within the comment, individual * and /
6270 characters may appear. An attempt to match C comments by applying the
6277 /* first comment */ not comment /* second comment */
6279 fails, because it matches the entire string owing to the greediness of
6282 However, if a quantifier is followed by a question mark, it ceases to
6283 be greedy, and instead matches the minimum number of times possible, so
6288 does the right thing with the C comments. The meaning of the various
6289 quantifiers is not otherwise changed, just the preferred number of
6290 matches. Do not confuse this use of question mark with its use as a
6291 quantifier in its own right. Because it has two uses, it can sometimes
6292 appear doubled, as in
6296 which matches one digit by preference, but can match two if that is the
6297 only way the rest of the pattern matches.
6299 If the PCRE_UNGREEDY option is set (an option that is not available in
6300 Perl), the quantifiers are not greedy by default, but individual ones
6301 can be made greedy by following them with a question mark. In other
6302 words, it inverts the default behaviour.
6304 When a parenthesized subpattern is quantified with a minimum repeat
6305 count that is greater than 1 or with a limited maximum, more memory is
6306 required for the compiled pattern, in proportion to the size of the
6309 If a pattern starts with .* or .{0,} and the PCRE_DOTALL option (equiv-
6310 alent to Perl's /s) is set, thus allowing the dot to match newlines,
6311 the pattern is implicitly anchored, because whatever follows will be
6312 tried against every character position in the subject string, so there
6313 is no point in retrying the overall match at any position after the
6314 first. PCRE normally treats such a pattern as though it were preceded
6317 In cases where it is known that the subject string contains no new-
6318 lines, it is worth setting PCRE_DOTALL in order to obtain this opti-
6319 mization, or alternatively using ^ to indicate anchoring explicitly.
6321 However, there are some cases where the optimization cannot be used.
6322 When .* is inside capturing parentheses that are the subject of a back
6323 reference elsewhere in the pattern, a match at the start may fail where
6324 a later one succeeds. Consider, for example:
6328 If the subject is "xyz123abc123" the match point is the fourth charac-
6329 ter. For this reason, such a pattern is not implicitly anchored.
6331 Another case where implicit anchoring is not applied is when the lead-
6332 ing .* is inside an atomic group. Once again, a match at the start may
6333 fail where a later one succeeds. Consider this pattern:
6337 It matches "ab" in the subject "aab". The use of the backtracking con-
6338 trol verbs (*PRUNE) and (*SKIP) also disable this optimization.
6340 When a capturing subpattern is repeated, the value captured is the sub-
6341 string that matched the final iteration. For example, after
6343 (tweedle[dume]{3}\s*)+
6345 has matched "tweedledum tweedledee" the value of the captured substring
6346 is "tweedledee". However, if there are nested capturing subpatterns,
6347 the corresponding captured values may have been set in previous itera-
6348 tions. For example, after
6352 matches "aba" the value of the second captured substring is "b".
6355 ATOMIC GROUPING AND POSSESSIVE QUANTIFIERS
6357 With both maximizing ("greedy") and minimizing ("ungreedy" or "lazy")
6358 repetition, failure of what follows normally causes the repeated item
6359 to be re-evaluated to see if a different number of repeats allows the
6360 rest of the pattern to match. Sometimes it is useful to prevent this,
6361 either to change the nature of the match, or to cause it fail earlier
6362 than it otherwise might, when the author of the pattern knows there is
6363 no point in carrying on.
6365 Consider, for example, the pattern \d+foo when applied to the subject
6370 After matching all 6 digits and then failing to match "foo", the normal
6371 action of the matcher is to try again with only 5 digits matching the
6372 \d+ item, and then with 4, and so on, before ultimately failing.
6373 "Atomic grouping" (a term taken from Jeffrey Friedl's book) provides
6374 the means for specifying that once a subpattern has matched, it is not
6375 to be re-evaluated in this way.
6377 If we use atomic grouping for the previous example, the matcher gives
6378 up immediately on failing to match "foo" the first time. The notation
6379 is a kind of special parenthesis, starting with (?> as in this example:
6383 This kind of parenthesis "locks up" the part of the pattern it con-
6384 tains once it has matched, and a failure further into the pattern is
6385 prevented from backtracking into it. Backtracking past it to previous
6386 items, however, works as normal.
6388 An alternative description is that a subpattern of this type matches
6389 the string of characters that an identical standalone pattern would
6390 match, if anchored at the current point in the subject string.
6392 Atomic grouping subpatterns are not capturing subpatterns. Simple cases
6393 such as the above example can be thought of as a maximizing repeat that
6394 must swallow everything it can. So, while both \d+ and \d+? are pre-
6395 pared to adjust the number of digits they match in order to make the
6396 rest of the pattern match, (?>\d+) can only match an entire sequence of
6399 Atomic groups in general can of course contain arbitrarily complicated
6400 subpatterns, and can be nested. However, when the subpattern for an
6401 atomic group is just a single repeated item, as in the example above, a
6402 simpler notation, called a "possessive quantifier" can be used. This
6403 consists of an additional + character following a quantifier. Using
6404 this notation, the previous example can be rewritten as
6408 Note that a possessive quantifier can be used with an entire group, for
6413 Possessive quantifiers are always greedy; the setting of the
6414 PCRE_UNGREEDY option is ignored. They are a convenient notation for the
6415 simpler forms of atomic group. However, there is no difference in the
6416 meaning of a possessive quantifier and the equivalent atomic group,
6417 though there may be a performance difference; possessive quantifiers
6418 should be slightly faster.
6420 The possessive quantifier syntax is an extension to the Perl 5.8 syn-
6421 tax. Jeffrey Friedl originated the idea (and the name) in the first
6422 edition of his book. Mike McCloskey liked it, so implemented it when he
6423 built Sun's Java package, and PCRE copied it from there. It ultimately
6424 found its way into Perl at release 5.10.
6426 PCRE has an optimization that automatically "possessifies" certain sim-
6427 ple pattern constructs. For example, the sequence A+B is treated as
6428 A++B because there is no point in backtracking into a sequence of A's
6431 When a pattern contains an unlimited repeat inside a subpattern that
6432 can itself be repeated an unlimited number of times, the use of an
6433 atomic group is the only way to avoid some failing matches taking a
6434 very long time indeed. The pattern
6438 matches an unlimited number of substrings that either consist of non-
6439 digits, or digits enclosed in <>, followed by either ! or ?. When it
6440 matches, it runs quickly. However, if it is applied to
6442 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
6444 it takes a long time before reporting failure. This is because the
6445 string can be divided between the internal \D+ repeat and the external
6446 * repeat in a large number of ways, and all have to be tried. (The
6447 example uses [!?] rather than a single character at the end, because
6448 both PCRE and Perl have an optimization that allows for fast failure
6449 when a single character is used. They remember the last single charac-
6450 ter that is required for a match, and fail early if it is not present
6451 in the string.) If the pattern is changed so that it uses an atomic
6454 ((?>\D+)|<\d+>)*[!?]
6456 sequences of non-digits cannot be broken, and failure happens quickly.
6461 Outside a character class, a backslash followed by a digit greater than
6462 0 (and possibly further digits) is a back reference to a capturing sub-
6463 pattern earlier (that is, to its left) in the pattern, provided there
6464 have been that many previous capturing left parentheses.
6466 However, if the decimal number following the backslash is less than 10,
6467 it is always taken as a back reference, and causes an error only if
6468 there are not that many capturing left parentheses in the entire pat-
6469 tern. In other words, the parentheses that are referenced need not be
6470 to the left of the reference for numbers less than 10. A "forward back
6471 reference" of this type can make sense when a repetition is involved
6472 and the subpattern to the right has participated in an earlier itera-
6475 It is not possible to have a numerical "forward back reference" to a
6476 subpattern whose number is 10 or more using this syntax because a
6477 sequence such as \50 is interpreted as a character defined in octal.
6478 See the subsection entitled "Non-printing characters" above for further
6479 details of the handling of digits following a backslash. There is no
6480 such problem when named parentheses are used. A back reference to any
6481 subpattern is possible using named parentheses (see below).
6483 Another way of avoiding the ambiguity inherent in the use of digits
6484 following a backslash is to use the \g escape sequence. This escape
6485 must be followed by an unsigned number or a negative number, optionally
6486 enclosed in braces. These examples are all identical:
6492 An unsigned number specifies an absolute reference without the ambigu-
6493 ity that is present in the older syntax. It is also useful when literal
6494 digits follow the reference. A negative number is a relative reference.
6495 Consider this example:
6499 The sequence \g{-1} is a reference to the most recently started captur-
6500 ing subpattern before \g, that is, is it equivalent to \2 in this exam-
6501 ple. Similarly, \g{-2} would be equivalent to \1. The use of relative
6502 references can be helpful in long patterns, and also in patterns that
6503 are created by joining together fragments that contain references
6506 A back reference matches whatever actually matched the capturing sub-
6507 pattern in the current subject string, rather than anything matching
6508 the subpattern itself (see "Subpatterns as subroutines" below for a way
6509 of doing that). So the pattern
6511 (sens|respons)e and \1ibility
6513 matches "sense and sensibility" and "response and responsibility", but
6514 not "sense and responsibility". If caseful matching is in force at the
6515 time of the back reference, the case of letters is relevant. For exam-
6520 matches "rah rah" and "RAH RAH", but not "RAH rah", even though the
6521 original capturing subpattern is matched caselessly.
6523 There are several different ways of writing back references to named
6524 subpatterns. The .NET syntax \k{name} and the Perl syntax \k<name> or
6525 \k'name' are supported, as is the Python syntax (?P=name). Perl 5.10's
6526 unified back reference syntax, in which \g can be used for both numeric
6527 and named references, is also supported. We could rewrite the above
6528 example in any of the following ways:
6530 (?<p1>(?i)rah)\s+\k<p1>
6531 (?'p1'(?i)rah)\s+\k{p1}
6532 (?P<p1>(?i)rah)\s+(?P=p1)
6533 (?<p1>(?i)rah)\s+\g{p1}
6535 A subpattern that is referenced by name may appear in the pattern
6536 before or after the reference.
6538 There may be more than one back reference to the same subpattern. If a
6539 subpattern has not actually been used in a particular match, any back
6540 references to it always fail by default. For example, the pattern
6544 always fails if it starts to match "a" rather than "bc". However, if
6545 the PCRE_JAVASCRIPT_COMPAT option is set at compile time, a back refer-
6546 ence to an unset value matches an empty string.
6548 Because there may be many capturing parentheses in a pattern, all dig-
6549 its following a backslash are taken as part of a potential back refer-
6550 ence number. If the pattern continues with a digit character, some
6551 delimiter must be used to terminate the back reference. If the
6552 PCRE_EXTENDED option is set, this can be white space. Otherwise, the
6553 \g{ syntax or an empty comment (see "Comments" below) can be used.
6555 Recursive back references
6557 A back reference that occurs inside the parentheses to which it refers
6558 fails when the subpattern is first used, so, for example, (a\1) never
6559 matches. However, such references can be useful inside repeated sub-
6560 patterns. For example, the pattern
6564 matches any number of "a"s and also "aba", "ababbaa" etc. At each iter-
6565 ation of the subpattern, the back reference matches the character
6566 string corresponding to the previous iteration. In order for this to
6567 work, the pattern must be such that the first iteration does not need
6568 to match the back reference. This can be done using alternation, as in
6569 the example above, or by a quantifier with a minimum of zero.
6571 Back references of this type cause the group that they reference to be
6572 treated as an atomic group. Once the whole group has been matched, a
6573 subsequent matching failure cannot cause backtracking into the middle
6579 An assertion is a test on the characters following or preceding the
6580 current matching point that does not actually consume any characters.
6581 The simple assertions coded as \b, \B, \A, \G, \Z, \z, ^ and $ are
6584 More complicated assertions are coded as subpatterns. There are two
6585 kinds: those that look ahead of the current position in the subject
6586 string, and those that look behind it. An assertion subpattern is
6587 matched in the normal way, except that it does not cause the current
6588 matching position to be changed.
6590 Assertion subpatterns are not capturing subpatterns. If such an asser-
6591 tion contains capturing subpatterns within it, these are counted for
6592 the purposes of numbering the capturing subpatterns in the whole pat-
6593 tern. However, substring capturing is carried out only for positive
6594 assertions. (Perl sometimes, but not always, does do capturing in nega-
6597 For compatibility with Perl, assertion subpatterns may be repeated;
6598 though it makes no sense to assert the same thing several times, the
6599 side effect of capturing parentheses may occasionally be useful. In
6600 practice, there only three cases:
6602 (1) If the quantifier is {0}, the assertion is never obeyed during
6603 matching. However, it may contain internal capturing parenthesized
6604 groups that are called from elsewhere via the subroutine mechanism.
6606 (2) If quantifier is {0,n} where n is greater than zero, it is treated
6607 as if it were {0,1}. At run time, the rest of the pattern match is
6608 tried with and without the assertion, the order depending on the greed-
6609 iness of the quantifier.
6611 (3) If the minimum repetition is greater than zero, the quantifier is
6612 ignored. The assertion is obeyed just once when encountered during
6615 Lookahead assertions
6617 Lookahead assertions start with (?= for positive assertions and (?! for
6618 negative assertions. For example,
6622 matches a word followed by a semicolon, but does not include the semi-
6623 colon in the match, and
6627 matches any occurrence of "foo" that is not followed by "bar". Note
6628 that the apparently similar pattern
6632 does not find an occurrence of "bar" that is preceded by something
6633 other than "foo"; it finds any occurrence of "bar" whatsoever, because
6634 the assertion (?!foo) is always true when the next three characters are
6635 "bar". A lookbehind assertion is needed to achieve the other effect.
6637 If you want to force a matching failure at some point in a pattern, the
6638 most convenient way to do it is with (?!) because an empty string
6639 always matches, so an assertion that requires there not to be an empty
6640 string must always fail. The backtracking control verb (*FAIL) or (*F)
6641 is a synonym for (?!).
6643 Lookbehind assertions
6645 Lookbehind assertions start with (?<= for positive assertions and (?<!
6646 for negative assertions. For example,
6650 does find an occurrence of "bar" that is not preceded by "foo". The
6651 contents of a lookbehind assertion are restricted such that all the
6652 strings it matches must have a fixed length. However, if there are sev-
6653 eral top-level alternatives, they do not all have to have the same
6662 causes an error at compile time. Branches that match different length
6663 strings are permitted only at the top level of a lookbehind assertion.
6664 This is an extension compared with Perl, which requires all branches to
6665 match the same length of string. An assertion such as
6669 is not permitted, because its single top-level branch can match two
6670 different lengths, but it is acceptable to PCRE if rewritten to use two
6675 In some cases, the escape sequence \K (see above) can be used instead
6676 of a lookbehind assertion to get round the fixed-length restriction.
6678 The implementation of lookbehind assertions is, for each alternative,
6679 to temporarily move the current position back by the fixed length and
6680 then try to match. If there are insufficient characters before the cur-
6681 rent position, the assertion fails.
6683 In a UTF mode, PCRE does not allow the \C escape (which matches a sin-
6684 gle data unit even in a UTF mode) to appear in lookbehind assertions,
6685 because it makes it impossible to calculate the length of the lookbe-
6686 hind. The \X and \R escapes, which can match different numbers of data
6687 units, are also not permitted.
6689 "Subroutine" calls (see below) such as (?2) or (?&X) are permitted in
6690 lookbehinds, as long as the subpattern matches a fixed-length string.
6691 Recursion, however, is not supported.
6693 Possessive quantifiers can be used in conjunction with lookbehind
6694 assertions to specify efficient matching of fixed-length strings at the
6695 end of subject strings. Consider a simple pattern such as
6699 when applied to a long string that does not match. Because matching
6700 proceeds from left to right, PCRE will look for each "a" in the subject
6701 and then see if what follows matches the rest of the pattern. If the
6702 pattern is specified as
6706 the initial .* matches the entire string at first, but when this fails
6707 (because there is no following "a"), it backtracks to match all but the
6708 last character, then all but the last two characters, and so on. Once
6709 again the search for "a" covers the entire string, from right to left,
6710 so we are no better off. However, if the pattern is written as
6714 there can be no backtracking for the .*+ item; it can match only the
6715 entire string. The subsequent lookbehind assertion does a single test
6716 on the last four characters. If it fails, the match fails immediately.
6717 For long strings, this approach makes a significant difference to the
6720 Using multiple assertions
6722 Several assertions (of any sort) may occur in succession. For example,
6724 (?<=\d{3})(?<!999)foo
6726 matches "foo" preceded by three digits that are not "999". Notice that
6727 each of the assertions is applied independently at the same point in
6728 the subject string. First there is a check that the previous three
6729 characters are all digits, and then there is a check that the same
6730 three characters are not "999". This pattern does not match "foo" pre-
6731 ceded by six characters, the first of which are digits and the last
6732 three of which are not "999". For example, it doesn't match "123abc-
6733 foo". A pattern to do that is
6735 (?<=\d{3}...)(?<!999)foo
6737 This time the first assertion looks at the preceding six characters,
6738 checking that the first three are digits, and then the second assertion
6739 checks that the preceding three characters are not "999".
6741 Assertions can be nested in any combination. For example,
6745 matches an occurrence of "baz" that is preceded by "bar" which in turn
6746 is not preceded by "foo", while
6748 (?<=\d{3}(?!999)...)foo
6750 is another pattern that matches "foo" preceded by three digits and any
6751 three characters that are not "999".
6754 CONDITIONAL SUBPATTERNS
6756 It is possible to cause the matching process to obey a subpattern con-
6757 ditionally or to choose between two alternative subpatterns, depending
6758 on the result of an assertion, or whether a specific capturing subpat-
6759 tern has already been matched. The two possible forms of conditional
6762 (?(condition)yes-pattern)
6763 (?(condition)yes-pattern|no-pattern)
6765 If the condition is satisfied, the yes-pattern is used; otherwise the
6766 no-pattern (if present) is used. If there are more than two alterna-
6767 tives in the subpattern, a compile-time error occurs. Each of the two
6768 alternatives may itself contain nested subpatterns of any form, includ-
6769 ing conditional subpatterns; the restriction to two alternatives
6770 applies only at the level of the condition. This pattern fragment is an
6771 example where the alternatives are complex:
6773 (?(1) (A|B|C) | (D | (?(2)E|F) | E) )
6776 There are four kinds of condition: references to subpatterns, refer-
6777 ences to recursion, a pseudo-condition called DEFINE, and assertions.
6779 Checking for a used subpattern by number
6781 If the text between the parentheses consists of a sequence of digits,
6782 the condition is true if a capturing subpattern of that number has pre-
6783 viously matched. If there is more than one capturing subpattern with
6784 the same number (see the earlier section about duplicate subpattern
6785 numbers), the condition is true if any of them have matched. An alter-
6786 native notation is to precede the digits with a plus or minus sign. In
6787 this case, the subpattern number is relative rather than absolute. The
6788 most recently opened parentheses can be referenced by (?(-1), the next
6789 most recent by (?(-2), and so on. Inside loops it can also make sense
6790 to refer to subsequent groups. The next parentheses to be opened can be
6791 referenced as (?(+1), and so on. (The value zero in any of these forms
6792 is not used; it provokes a compile-time error.)
6794 Consider the following pattern, which contains non-significant white
6795 space to make it more readable (assume the PCRE_EXTENDED option) and to
6796 divide it into three parts for ease of discussion:
6798 ( \( )? [^()]+ (?(1) \) )
6800 The first part matches an optional opening parenthesis, and if that
6801 character is present, sets it as the first captured substring. The sec-
6802 ond part matches one or more characters that are not parentheses. The
6803 third part is a conditional subpattern that tests whether or not the
6804 first set of parentheses matched. If they did, that is, if subject
6805 started with an opening parenthesis, the condition is true, and so the
6806 yes-pattern is executed and a closing parenthesis is required. Other-
6807 wise, since no-pattern is not present, the subpattern matches nothing.
6808 In other words, this pattern matches a sequence of non-parentheses,
6809 optionally enclosed in parentheses.
6811 If you were embedding this pattern in a larger one, you could use a
6814 ...other stuff... ( \( )? [^()]+ (?(-1) \) ) ...
6816 This makes the fragment independent of the parentheses in the larger
6819 Checking for a used subpattern by name
6821 Perl uses the syntax (?(<name>)...) or (?('name')...) to test for a
6822 used subpattern by name. For compatibility with earlier versions of
6823 PCRE, which had this facility before Perl, the syntax (?(name)...) is
6826 Rewriting the above example to use a named subpattern gives this:
6828 (?<OPEN> \( )? [^()]+ (?(<OPEN>) \) )
6830 If the name used in a condition of this kind is a duplicate, the test
6831 is applied to all subpatterns of the same name, and is true if any one
6832 of them has matched.
6834 Checking for pattern recursion
6836 If the condition is the string (R), and there is no subpattern with the
6837 name R, the condition is true if a recursive call to the whole pattern
6838 or any subpattern has been made. If digits or a name preceded by amper-
6839 sand follow the letter R, for example:
6841 (?(R3)...) or (?(R&name)...)
6843 the condition is true if the most recent recursion is into a subpattern
6844 whose number or name is given. This condition does not check the entire
6845 recursion stack. If the name used in a condition of this kind is a
6846 duplicate, the test is applied to all subpatterns of the same name, and
6847 is true if any one of them is the most recent recursion.
6849 At "top level", all these recursion test conditions are false. The
6850 syntax for recursive patterns is described below.
6852 Defining subpatterns for use by reference only
6854 If the condition is the string (DEFINE), and there is no subpattern
6855 with the name DEFINE, the condition is always false. In this case,
6856 there may be only one alternative in the subpattern. It is always
6857 skipped if control reaches this point in the pattern; the idea of
6858 DEFINE is that it can be used to define subroutines that can be refer-
6859 enced from elsewhere. (The use of subroutines is described below.) For
6860 example, a pattern to match an IPv4 address such as "192.168.23.245"
6861 could be written like this (ignore white space and line breaks):
6863 (?(DEFINE) (?<byte> 2[0-4]\d | 25[0-5] | 1\d\d | [1-9]?\d) )
6864 \b (?&byte) (\.(?&byte)){3} \b
6866 The first part of the pattern is a DEFINE group inside which a another
6867 group named "byte" is defined. This matches an individual component of
6868 an IPv4 address (a number less than 256). When matching takes place,
6869 this part of the pattern is skipped because DEFINE acts like a false
6870 condition. The rest of the pattern uses references to the named group
6871 to match the four dot-separated components of an IPv4 address, insist-
6872 ing on a word boundary at each end.
6874 Assertion conditions
6876 If the condition is not in any of the above formats, it must be an
6877 assertion. This may be a positive or negative lookahead or lookbehind
6878 assertion. Consider this pattern, again containing non-significant
6879 white space, and with the two alternatives on the second line:
6882 \d{2}-[a-z]{3}-\d{2} | \d{2}-\d{2}-\d{2} )
6884 The condition is a positive lookahead assertion that matches an
6885 optional sequence of non-letters followed by a letter. In other words,
6886 it tests for the presence of at least one letter in the subject. If a
6887 letter is found, the subject is matched against the first alternative;
6888 otherwise it is matched against the second. This pattern matches
6889 strings in one of the two forms dd-aaa-dd or dd-dd-dd, where aaa are
6890 letters and dd are digits.
6895 There are two ways of including comments in patterns that are processed
6896 by PCRE. In both cases, the start of the comment must not be in a char-
6897 acter class, nor in the middle of any other sequence of related charac-
6898 ters such as (?: or a subpattern name or number. The characters that
6899 make up a comment play no part in the pattern matching.
6901 The sequence (?# marks the start of a comment that continues up to the
6902 next closing parenthesis. Nested parentheses are not permitted. If the
6903 PCRE_EXTENDED option is set, an unescaped # character also introduces a
6904 comment, which in this case continues to immediately after the next
6905 newline character or character sequence in the pattern. Which charac-
6906 ters are interpreted as newlines is controlled by the options passed to
6907 a compiling function or by a special sequence at the start of the pat-
6908 tern, as described in the section entitled "Newline conventions" above.
6909 Note that the end of this type of comment is a literal newline sequence
6910 in the pattern; escape sequences that happen to represent a newline do
6911 not count. For example, consider this pattern when PCRE_EXTENDED is
6912 set, and the default newline convention is in force:
6914 abc #comment \n still comment
6916 On encountering the # character, pcre_compile() skips along, looking
6917 for a newline in the pattern. The sequence \n is still literal at this
6918 stage, so it does not terminate the comment. Only an actual character
6919 with the code value 0x0a (the default newline) does so.
6924 Consider the problem of matching a string in parentheses, allowing for
6925 unlimited nested parentheses. Without the use of recursion, the best
6926 that can be done is to use a pattern that matches up to some fixed
6927 depth of nesting. It is not possible to handle an arbitrary nesting
6930 For some time, Perl has provided a facility that allows regular expres-
6931 sions to recurse (amongst other things). It does this by interpolating
6932 Perl code in the expression at run time, and the code can refer to the
6933 expression itself. A Perl pattern using code interpolation to solve the
6934 parentheses problem can be created like this:
6936 $re = qr{\( (?: (?>[^()]+) | (?p{$re}) )* \)}x;
6938 The (?p{...}) item interpolates Perl code at run time, and in this case
6939 refers recursively to the pattern in which it appears.
6941 Obviously, PCRE cannot support the interpolation of Perl code. Instead,
6942 it supports special syntax for recursion of the entire pattern, and
6943 also for individual subpattern recursion. After its introduction in
6944 PCRE and Python, this kind of recursion was subsequently introduced
6945 into Perl at release 5.10.
6947 A special item that consists of (? followed by a number greater than
6948 zero and a closing parenthesis is a recursive subroutine call of the
6949 subpattern of the given number, provided that it occurs inside that
6950 subpattern. (If not, it is a non-recursive subroutine call, which is
6951 described in the next section.) The special item (?R) or (?0) is a
6952 recursive call of the entire regular expression.
6954 This PCRE pattern solves the nested parentheses problem (assume the
6955 PCRE_EXTENDED option is set so that white space is ignored):
6957 \( ( [^()]++ | (?R) )* \)
6959 First it matches an opening parenthesis. Then it matches any number of
6960 substrings which can either be a sequence of non-parentheses, or a
6961 recursive match of the pattern itself (that is, a correctly parenthe-
6962 sized substring). Finally there is a closing parenthesis. Note the use
6963 of a possessive quantifier to avoid backtracking into sequences of non-
6966 If this were part of a larger pattern, you would not want to recurse
6967 the entire pattern, so instead you could use this:
6969 ( \( ( [^()]++ | (?1) )* \) )
6971 We have put the pattern into parentheses, and caused the recursion to
6972 refer to them instead of the whole pattern.
6974 In a larger pattern, keeping track of parenthesis numbers can be
6975 tricky. This is made easier by the use of relative references. Instead
6976 of (?1) in the pattern above you can write (?-2) to refer to the second
6977 most recently opened parentheses preceding the recursion. In other
6978 words, a negative number counts capturing parentheses leftwards from
6979 the point at which it is encountered.
6981 It is also possible to refer to subsequently opened parentheses, by
6982 writing references such as (?+2). However, these cannot be recursive
6983 because the reference is not inside the parentheses that are refer-
6984 enced. They are always non-recursive subroutine calls, as described in
6987 An alternative approach is to use named parentheses instead. The Perl
6988 syntax for this is (?&name); PCRE's earlier syntax (?P>name) is also
6989 supported. We could rewrite the above example as follows:
6991 (?<pn> \( ( [^()]++ | (?&pn) )* \) )
6993 If there is more than one subpattern with the same name, the earliest
6996 This particular example pattern that we have been looking at contains
6997 nested unlimited repeats, and so the use of a possessive quantifier for
6998 matching strings of non-parentheses is important when applying the pat-
6999 tern to strings that do not match. For example, when this pattern is
7002 (aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa()
7004 it yields "no match" quickly. However, if a possessive quantifier is
7005 not used, the match runs for a very long time indeed because there are
7006 so many different ways the + and * repeats can carve up the subject,
7007 and all have to be tested before failure can be reported.
7009 At the end of a match, the values of capturing parentheses are those
7010 from the outermost level. If you want to obtain intermediate values, a
7011 callout function can be used (see below and the pcrecallout documenta-
7012 tion). If the pattern above is matched against
7016 the value for the inner capturing parentheses (numbered 2) is "ef",
7017 which is the last value taken on at the top level. If a capturing sub-
7018 pattern is not matched at the top level, its final captured value is
7019 unset, even if it was (temporarily) set at a deeper level during the
7022 If there are more than 15 capturing parentheses in a pattern, PCRE has
7023 to obtain extra memory to store data during a recursion, which it does
7024 by using pcre_malloc, freeing it via pcre_free afterwards. If no memory
7025 can be obtained, the match fails with the PCRE_ERROR_NOMEMORY error.
7027 Do not confuse the (?R) item with the condition (R), which tests for
7028 recursion. Consider this pattern, which matches text in angle brack-
7029 ets, allowing for arbitrary nesting. Only digits are allowed in nested
7030 brackets (that is, when recursing), whereas any characters are permit-
7031 ted at the outer level.
7033 < (?: (?(R) \d++ | [^<>]*+) | (?R)) * >
7035 In this pattern, (?(R) is the start of a conditional subpattern, with
7036 two different alternatives for the recursive and non-recursive cases.
7037 The (?R) item is the actual recursive call.
7039 Differences in recursion processing between PCRE and Perl
7041 Recursion processing in PCRE differs from Perl in two important ways.
7042 In PCRE (like Python, but unlike Perl), a recursive subpattern call is
7043 always treated as an atomic group. That is, once it has matched some of
7044 the subject string, it is never re-entered, even if it contains untried
7045 alternatives and there is a subsequent matching failure. This can be
7046 illustrated by the following pattern, which purports to match a palin-
7047 dromic string that contains an odd number of characters (for example,
7048 "a", "aba", "abcba", "abcdcba"):
7052 The idea is that it either matches a single character, or two identical
7053 characters surrounding a sub-palindrome. In Perl, this pattern works;
7054 in PCRE it does not if the pattern is longer than three characters.
7055 Consider the subject string "abcba":
7057 At the top level, the first character is matched, but as it is not at
7058 the end of the string, the first alternative fails; the second alterna-
7059 tive is taken and the recursion kicks in. The recursive call to subpat-
7060 tern 1 successfully matches the next character ("b"). (Note that the
7061 beginning and end of line tests are not part of the recursion).
7063 Back at the top level, the next character ("c") is compared with what
7064 subpattern 2 matched, which was "a". This fails. Because the recursion
7065 is treated as an atomic group, there are now no backtracking points,
7066 and so the entire match fails. (Perl is able, at this point, to re-
7067 enter the recursion and try the second alternative.) However, if the
7068 pattern is written with the alternatives in the other order, things are
7073 This time, the recursing alternative is tried first, and continues to
7074 recurse until it runs out of characters, at which point the recursion
7075 fails. But this time we do have another alternative to try at the
7076 higher level. That is the big difference: in the previous case the
7077 remaining alternative is at a deeper recursion level, which PCRE cannot
7080 To change the pattern so that it matches all palindromic strings, not
7081 just those with an odd number of characters, it is tempting to change
7082 the pattern to this:
7086 Again, this works in Perl, but not in PCRE, and for the same reason.
7087 When a deeper recursion has matched a single character, it cannot be
7088 entered again in order to match an empty string. The solution is to
7089 separate the two cases, and write out the odd and even cases as alter-
7090 natives at the higher level:
7092 ^(?:((.)(?1)\2|)|((.)(?3)\4|.))
7094 If you want to match typical palindromic phrases, the pattern has to
7095 ignore all non-word characters, which can be done like this:
7097 ^\W*+(?:((.)\W*+(?1)\W*+\2|)|((.)\W*+(?3)\W*+\4|\W*+.\W*+))\W*+$
7099 If run with the PCRE_CASELESS option, this pattern matches phrases such
7100 as "A man, a plan, a canal: Panama!" and it works well in both PCRE and
7101 Perl. Note the use of the possessive quantifier *+ to avoid backtrack-
7102 ing into sequences of non-word characters. Without this, PCRE takes a
7103 great deal longer (ten times or more) to match typical phrases, and
7104 Perl takes so long that you think it has gone into a loop.
7106 WARNING: The palindrome-matching patterns above work only if the sub-
7107 ject string does not start with a palindrome that is shorter than the
7108 entire string. For example, although "abcba" is correctly matched, if
7109 the subject is "ababa", PCRE finds the palindrome "aba" at the start,
7110 then fails at top level because the end of the string does not follow.
7111 Once again, it cannot jump back into the recursion to try other alter-
7112 natives, so the entire match fails.
7114 The second way in which PCRE and Perl differ in their recursion pro-
7115 cessing is in the handling of captured values. In Perl, when a subpat-
7116 tern is called recursively or as a subpattern (see the next section),
7117 it has no access to any values that were captured outside the recur-
7118 sion, whereas in PCRE these values can be referenced. Consider this
7123 In PCRE, this pattern matches "bab". The first capturing parentheses
7124 match "b", then in the second group, when the back reference \1 fails
7125 to match "b", the second alternative matches "a" and then recurses. In
7126 the recursion, \1 does now match "b" and so the whole match succeeds.
7127 In Perl, the pattern fails to match because inside the recursive call
7128 \1 cannot access the externally set value.
7131 SUBPATTERNS AS SUBROUTINES
7133 If the syntax for a recursive subpattern call (either by number or by
7134 name) is used outside the parentheses to which it refers, it operates
7135 like a subroutine in a programming language. The called subpattern may
7136 be defined before or after the reference. A numbered reference can be
7137 absolute or relative, as in these examples:
7139 (...(absolute)...)...(?2)...
7140 (...(relative)...)...(?-1)...
7141 (...(?+1)...(relative)...
7143 An earlier example pointed out that the pattern
7145 (sens|respons)e and \1ibility
7147 matches "sense and sensibility" and "response and responsibility", but
7148 not "sense and responsibility". If instead the pattern
7150 (sens|respons)e and (?1)ibility
7152 is used, it does match "sense and responsibility" as well as the other
7153 two strings. Another example is given in the discussion of DEFINE
7156 All subroutine calls, whether recursive or not, are always treated as
7157 atomic groups. That is, once a subroutine has matched some of the sub-
7158 ject string, it is never re-entered, even if it contains untried alter-
7159 natives and there is a subsequent matching failure. Any capturing
7160 parentheses that are set during the subroutine call revert to their
7161 previous values afterwards.
7163 Processing options such as case-independence are fixed when a subpat-
7164 tern is defined, so if it is used as a subroutine, such options cannot
7165 be changed for different calls. For example, consider this pattern:
7169 It matches "abcabc". It does not match "abcABC" because the change of
7170 processing option does not affect the called subpattern.
7173 ONIGURUMA SUBROUTINE SYNTAX
7175 For compatibility with Oniguruma, the non-Perl syntax \g followed by a
7176 name or a number enclosed either in angle brackets or single quotes, is
7177 an alternative syntax for referencing a subpattern as a subroutine,
7178 possibly recursively. Here are two of the examples used above, rewrit-
7179 ten using this syntax:
7181 (?<pn> \( ( (?>[^()]+) | \g<pn> )* \) )
7182 (sens|respons)e and \g'1'ibility
7184 PCRE supports an extension to Oniguruma: if a number is preceded by a
7185 plus or a minus sign it is taken as a relative reference. For example:
7189 Note that \g{...} (Perl syntax) and \g<...> (Oniguruma syntax) are not
7190 synonymous. The former is a back reference; the latter is a subroutine
7196 Perl has a feature whereby using the sequence (?{...}) causes arbitrary
7197 Perl code to be obeyed in the middle of matching a regular expression.
7198 This makes it possible, amongst other things, to extract different sub-
7199 strings that match the same pair of parentheses when there is a repeti-
7202 PCRE provides a similar feature, but of course it cannot obey arbitrary
7203 Perl code. The feature is called "callout". The caller of PCRE provides
7204 an external function by putting its entry point in the global variable
7205 pcre_callout (8-bit library) or pcre[16|32]_callout (16-bit or 32-bit
7206 library). By default, this variable contains NULL, which disables all
7209 Within a regular expression, (?C) indicates the points at which the
7210 external function is to be called. If you want to identify different
7211 callout points, you can put a number less than 256 after the letter C.
7212 The default value is zero. For example, this pattern has two callout
7217 If the PCRE_AUTO_CALLOUT flag is passed to a compiling function, call-
7218 outs are automatically installed before each item in the pattern. They
7219 are all numbered 255. If there is a conditional group in the pattern
7220 whose condition is an assertion, an additional callout is inserted just
7221 before the condition. An explicit callout may also be set at this posi-
7222 tion, as in this example:
7224 (?(?C9)(?=a)abc|def)
7226 Note that this applies only to assertion conditions, not to other types
7229 During matching, when PCRE reaches a callout point, the external func-
7230 tion is called. It is provided with the number of the callout, the
7231 position in the pattern, and, optionally, one item of data originally
7232 supplied by the caller of the matching function. The callout function
7233 may cause matching to proceed, to backtrack, or to fail altogether.
7235 By default, PCRE implements a number of optimizations at compile time
7236 and matching time, and one side-effect is that sometimes callouts are
7237 skipped. If you need all possible callouts to happen, you need to set
7238 options that disable the relevant optimizations. More details, and a
7239 complete description of the interface to the callout function, are
7240 given in the pcrecallout documentation.
7243 BACKTRACKING CONTROL
7245 Perl 5.10 introduced a number of "Special Backtracking Control Verbs",
7246 which are still described in the Perl documentation as "experimental
7247 and subject to change or removal in a future version of Perl". It goes
7248 on to say: "Their usage in production code should be noted to avoid
7249 problems during upgrades." The same remarks apply to the PCRE features
7250 described in this section.
7252 The new verbs make use of what was previously invalid syntax: an open-
7253 ing parenthesis followed by an asterisk. They are generally of the form
7254 (*VERB) or (*VERB:NAME). Some may take either form, possibly behaving
7255 differently depending on whether or not a name is present. A name is
7256 any sequence of characters that does not include a closing parenthesis.
7257 The maximum length of name is 255 in the 8-bit library and 65535 in the
7258 16-bit and 32-bit libraries. If the name is empty, that is, if the
7259 closing parenthesis immediately follows the colon, the effect is as if
7260 the colon were not there. Any number of these verbs may occur in a
7263 Since these verbs are specifically related to backtracking, most of
7264 them can be used only when the pattern is to be matched using one of
7265 the traditional matching functions, because these use a backtracking
7266 algorithm. With the exception of (*FAIL), which behaves like a failing
7267 negative assertion, the backtracking control verbs cause an error if
7268 encountered by a DFA matching function.
7270 The behaviour of these verbs in repeated groups, assertions, and in
7271 subpatterns called as subroutines (whether or not recursively) is docu-
7274 Optimizations that affect backtracking verbs
7276 PCRE contains some optimizations that are used to speed up matching by
7277 running some checks at the start of each match attempt. For example, it
7278 may know the minimum length of matching subject, or that a particular
7279 character must be present. When one of these optimizations bypasses the
7280 running of a match, any included backtracking verbs will not, of
7281 course, be processed. You can suppress the start-of-match optimizations
7282 by setting the PCRE_NO_START_OPTIMIZE option when calling pcre_com-
7283 pile() or pcre_exec(), or by starting the pattern with (*NO_START_OPT).
7284 There is more discussion of this option in the section entitled "Option
7285 bits for pcre_exec()" in the pcreapi documentation.
7287 Experiments with Perl suggest that it too has similar optimizations,
7288 sometimes leading to anomalous results.
7290 Verbs that act immediately
7292 The following verbs act as soon as they are encountered. They may not
7293 be followed by a name.
7297 This verb causes the match to end successfully, skipping the remainder
7298 of the pattern. However, when it is inside a subpattern that is called
7299 as a subroutine, only that subpattern is ended successfully. Matching
7300 then continues at the outer level. If (*ACCEPT) in triggered in a posi-
7301 tive assertion, the assertion succeeds; in a negative assertion, the
7304 If (*ACCEPT) is inside capturing parentheses, the data so far is cap-
7307 A((?:A|B(*ACCEPT)|C)D)
7309 This matches "AB", "AAD", or "ACD"; when it matches "AB", "B" is cap-
7310 tured by the outer parentheses.
7314 This verb causes a matching failure, forcing backtracking to occur. It
7315 is equivalent to (?!) but easier to read. The Perl documentation notes
7316 that it is probably useful only when combined with (?{}) or (??{}).
7317 Those are, of course, Perl features that are not present in PCRE. The
7318 nearest equivalent is the callout feature, as for example in this pat-
7323 A match with the string "aaaa" always fails, but the callout is taken
7324 before each backtrack happens (in this example, 10 times).
7326 Recording which path was taken
7328 There is one verb whose main purpose is to track how a match was
7329 arrived at, though it also has a secondary use in conjunction with
7330 advancing the match starting point (see (*SKIP) below).
7332 (*MARK:NAME) or (*:NAME)
7334 A name is always required with this verb. There may be as many
7335 instances of (*MARK) as you like in a pattern, and their names do not
7338 When a match succeeds, the name of the last-encountered (*MARK:NAME),
7339 (*PRUNE:NAME), or (*THEN:NAME) on the matching path is passed back to
7340 the caller as described in the section entitled "Extra data for
7341 pcre_exec()" in the pcreapi documentation. Here is an example of
7342 pcretest output, where the /K modifier requests the retrieval and out-
7343 putting of (*MARK) data:
7345 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
7353 The (*MARK) name is tagged with "MK:" in this output, and in this exam-
7354 ple it indicates which of the two alternatives matched. This is a more
7355 efficient way of obtaining this information than putting each alterna-
7356 tive in its own capturing parentheses.
7358 If a verb with a name is encountered in a positive assertion that is
7359 true, the name is recorded and passed back if it is the last-encoun-
7360 tered. This does not happen for negative assertions or failing positive
7363 After a partial match or a failed match, the last encountered name in
7364 the entire match process is returned. For example:
7366 re> /X(*MARK:A)Y|X(*MARK:B)Z/K
7370 Note that in this unanchored example the mark is retained from the
7371 match attempt that started at the letter "X" in the subject. Subsequent
7372 match attempts starting at "P" and then with an empty string do not get
7373 as far as the (*MARK) item, but nevertheless do not reset it.
7375 If you are interested in (*MARK) values after failed matches, you
7376 should probably set the PCRE_NO_START_OPTIMIZE option (see above) to
7377 ensure that the match is always attempted.
7379 Verbs that act after backtracking
7381 The following verbs do nothing when they are encountered. Matching con-
7382 tinues with what follows, but if there is no subsequent match, causing
7383 a backtrack to the verb, a failure is forced. That is, backtracking
7384 cannot pass to the left of the verb. However, when one of these verbs
7385 appears inside an atomic group or an assertion that is true, its effect
7386 is confined to that group, because once the group has been matched,
7387 there is never any backtracking into it. In this situation, backtrack-
7388 ing can "jump back" to the left of the entire atomic group or asser-
7389 tion. (Remember also, as stated above, that this localization also
7390 applies in subroutine calls.)
7392 These verbs differ in exactly what kind of failure occurs when back-
7393 tracking reaches them. The behaviour described below is what happens
7394 when the verb is not in a subroutine or an assertion. Subsequent sec-
7395 tions cover these special cases.
7399 This verb, which may not be followed by a name, causes the whole match
7400 to fail outright if there is a later matching failure that causes back-
7401 tracking to reach it. Even if the pattern is unanchored, no further
7402 attempts to find a match by advancing the starting point take place. If
7403 (*COMMIT) is the only backtracking verb that is encountered, once it
7404 has been passed pcre_exec() is committed to finding a match at the cur-
7405 rent starting point, or not at all. For example:
7409 This matches "xxaab" but not "aacaab". It can be thought of as a kind
7410 of dynamic anchor, or "I've started, so I must finish." The name of the
7411 most recently passed (*MARK) in the path is passed back when (*COMMIT)
7412 forces a match failure.
7414 If there is more than one backtracking verb in a pattern, a different
7415 one that follows (*COMMIT) may be triggered first, so merely passing
7416 (*COMMIT) during a match does not always guarantee that a match must be
7417 at this starting point.
7419 Note that (*COMMIT) at the start of a pattern is not the same as an
7420 anchor, unless PCRE's start-of-match optimizations are turned off, as
7421 shown in this output from pcretest:
7429 For this pattern, PCRE knows that any match must start with "a", so the
7430 optimization skips along the subject to "a" before applying the pattern
7431 to the first set of data. The match attempt then succeeds. In the sec-
7432 ond set of data, the escape sequence \Y is interpreted by the pcretest
7433 program. It causes the PCRE_NO_START_OPTIMIZE option to be set when
7434 pcre_exec() is called. This disables the optimization that skips along
7435 to the first character. The pattern is now applied starting at "x", and
7436 so the (*COMMIT) causes the match to fail without trying any other
7439 (*PRUNE) or (*PRUNE:NAME)
7441 This verb causes the match to fail at the current starting position in
7442 the subject if there is a later matching failure that causes backtrack-
7443 ing to reach it. If the pattern is unanchored, the normal "bumpalong"
7444 advance to the next starting character then happens. Backtracking can
7445 occur as usual to the left of (*PRUNE), before it is reached, or when
7446 matching to the right of (*PRUNE), but if there is no match to the
7447 right, backtracking cannot cross (*PRUNE). In simple cases, the use of
7448 (*PRUNE) is just an alternative to an atomic group or possessive quan-
7449 tifier, but there are some uses of (*PRUNE) that cannot be expressed in
7450 any other way. In an anchored pattern (*PRUNE) has the same effect as
7453 The behaviour of (*PRUNE:NAME) is the not the same as
7454 (*MARK:NAME)(*PRUNE). It is like (*MARK:NAME) in that the name is
7455 remembered for passing back to the caller. However, (*SKIP:NAME)
7456 searches only for names set with (*MARK).
7460 This verb, when given without a name, is like (*PRUNE), except that if
7461 the pattern is unanchored, the "bumpalong" advance is not to the next
7462 character, but to the position in the subject where (*SKIP) was encoun-
7463 tered. (*SKIP) signifies that whatever text was matched leading up to
7464 it cannot be part of a successful match. Consider:
7468 If the subject is "aaaac...", after the first match attempt fails
7469 (starting at the first character in the string), the starting point
7470 skips on to start the next attempt at "c". Note that a possessive quan-
7471 tifer does not have the same effect as this example; although it would
7472 suppress backtracking during the first match attempt, the second
7473 attempt would start at the second character instead of skipping on to
7478 When (*SKIP) has an associated name, its behaviour is modified. When it
7479 is triggered, the previous path through the pattern is searched for the
7480 most recent (*MARK) that has the same name. If one is found, the
7481 "bumpalong" advance is to the subject position that corresponds to that
7482 (*MARK) instead of to where (*SKIP) was encountered. If no (*MARK) with
7483 a matching name is found, the (*SKIP) is ignored.
7485 Note that (*SKIP:NAME) searches only for names set by (*MARK:NAME). It
7486 ignores names that are set by (*PRUNE:NAME) or (*THEN:NAME).
7488 (*THEN) or (*THEN:NAME)
7490 This verb causes a skip to the next innermost alternative when back-
7491 tracking reaches it. That is, it cancels any further backtracking
7492 within the current alternative. Its name comes from the observation
7493 that it can be used for a pattern-based if-then-else block:
7495 ( COND1 (*THEN) FOO | COND2 (*THEN) BAR | COND3 (*THEN) BAZ ) ...
7497 If the COND1 pattern matches, FOO is tried (and possibly further items
7498 after the end of the group if FOO succeeds); on failure, the matcher
7499 skips to the second alternative and tries COND2, without backtracking
7500 into COND1. If that succeeds and BAR fails, COND3 is tried. If subse-
7501 quently BAZ fails, there are no more alternatives, so there is a back-
7502 track to whatever came before the entire group. If (*THEN) is not
7503 inside an alternation, it acts like (*PRUNE).
7505 The behaviour of (*THEN:NAME) is the not the same as
7506 (*MARK:NAME)(*THEN). It is like (*MARK:NAME) in that the name is
7507 remembered for passing back to the caller. However, (*SKIP:NAME)
7508 searches only for names set with (*MARK).
7510 A subpattern that does not contain a | character is just a part of the
7511 enclosing alternative; it is not a nested alternation with only one
7512 alternative. The effect of (*THEN) extends beyond such a subpattern to
7513 the enclosing alternative. Consider this pattern, where A, B, etc. are
7514 complex pattern fragments that do not contain any | characters at this
7519 If A and B are matched, but there is a failure in C, matching does not
7520 backtrack into A; instead it moves to the next alternative, that is, D.
7521 However, if the subpattern containing (*THEN) is given an alternative,
7522 it behaves differently:
7524 A (B(*THEN)C | (*FAIL)) | D
7526 The effect of (*THEN) is now confined to the inner subpattern. After a
7527 failure in C, matching moves to (*FAIL), which causes the whole subpat-
7528 tern to fail because there are no more alternatives to try. In this
7529 case, matching does now backtrack into A.
7531 Note that a conditional subpattern is not considered as having two
7532 alternatives, because only one is ever used. In other words, the |
7533 character in a conditional subpattern has a different meaning. Ignoring
7534 white space, consider:
7536 ^.*? (?(?=a) a | b(*THEN)c )
7538 If the subject is "ba", this pattern does not match. Because .*? is
7539 ungreedy, it initially matches zero characters. The condition (?=a)
7540 then fails, the character "b" is matched, but "c" is not. At this
7541 point, matching does not backtrack to .*? as might perhaps be expected
7542 from the presence of the | character. The conditional subpattern is
7543 part of the single alternative that comprises the whole pattern, and so
7544 the match fails. (If there was a backtrack into .*?, allowing it to
7545 match "b", the match would succeed.)
7547 The verbs just described provide four different "strengths" of control
7548 when subsequent matching fails. (*THEN) is the weakest, carrying on the
7549 match at the next alternative. (*PRUNE) comes next, failing the match
7550 at the current starting position, but allowing an advance to the next
7551 character (for an unanchored pattern). (*SKIP) is similar, except that
7552 the advance may be more than one character. (*COMMIT) is the strongest,
7553 causing the entire match to fail.
7555 More than one backtracking verb
7557 If more than one backtracking verb is present in a pattern, the one
7558 that is backtracked onto first acts. For example, consider this pat-
7559 tern, where A, B, etc. are complex pattern fragments:
7561 (A(*COMMIT)B(*THEN)C|ABD)
7563 If A matches but B fails, the backtrack to (*COMMIT) causes the entire
7564 match to fail. However, if A and B match, but C fails, the backtrack to
7565 (*THEN) causes the next alternative (ABD) to be tried. This behaviour
7566 is consistent, but is not always the same as Perl's. It means that if
7567 two or more backtracking verbs appear in succession, all the the last
7568 of them has no effect. Consider this example:
7570 ...(*COMMIT)(*PRUNE)...
7572 If there is a matching failure to the right, backtracking onto (*PRUNE)
7573 causes it to be triggered, and its action is taken. There can never be
7574 a backtrack onto (*COMMIT).
7576 Backtracking verbs in repeated groups
7578 PCRE differs from Perl in its handling of backtracking verbs in
7579 repeated groups. For example, consider:
7583 If the subject is "abac", Perl matches, but PCRE fails because the
7584 (*COMMIT) in the second repeat of the group acts.
7586 Backtracking verbs in assertions
7588 (*FAIL) in an assertion has its normal effect: it forces an immediate
7591 (*ACCEPT) in a positive assertion causes the assertion to succeed with-
7592 out any further processing. In a negative assertion, (*ACCEPT) causes
7593 the assertion to fail without any further processing.
7595 The other backtracking verbs are not treated specially if they appear
7596 in a positive assertion. In particular, (*THEN) skips to the next
7597 alternative in the innermost enclosing group that has alternations,
7598 whether or not this is within the assertion.
7600 Negative assertions are, however, different, in order to ensure that
7601 changing a positive assertion into a negative assertion changes its
7602 result. Backtracking into (*COMMIT), (*SKIP), or (*PRUNE) causes a neg-
7603 ative assertion to be true, without considering any further alternative
7604 branches in the assertion. Backtracking into (*THEN) causes it to skip
7605 to the next enclosing alternative within the assertion (the normal be-
7606 haviour), but if the assertion does not have such an alternative,
7607 (*THEN) behaves like (*PRUNE).
7609 Backtracking verbs in subroutines
7611 These behaviours occur whether or not the subpattern is called recur-
7612 sively. Perl's treatment of subroutines is different in some cases.
7614 (*FAIL) in a subpattern called as a subroutine has its normal effect:
7615 it forces an immediate backtrack.
7617 (*ACCEPT) in a subpattern called as a subroutine causes the subroutine
7618 match to succeed without any further processing. Matching then contin-
7619 ues after the subroutine call.
7621 (*COMMIT), (*SKIP), and (*PRUNE) in a subpattern called as a subroutine
7622 cause the subroutine match to fail.
7624 (*THEN) skips to the next alternative in the innermost enclosing group
7625 within the subpattern that has alternatives. If there is no such group
7626 within the subpattern, (*THEN) causes the subroutine match to fail.
7631 pcreapi(3), pcrecallout(3), pcrematching(3), pcresyntax(3), pcre(3),
7632 pcre16(3), pcre32(3).
7638 University Computing Service
7639 Cambridge CB2 3QH, England.
7644 Last updated: 08 January 2014
7645 Copyright (c) 1997-2014 University of Cambridge.
7646 ------------------------------------------------------------------------------
7649 PCRESYNTAX(3) Library Functions Manual PCRESYNTAX(3)
7654 PCRE - Perl-compatible regular expressions
7656 PCRE REGULAR EXPRESSION SYNTAX SUMMARY
7658 The full syntax and semantics of the regular expressions that are sup-
7659 ported by PCRE are described in the pcrepattern documentation. This
7660 document contains a quick-reference summary of the syntax.
7665 \x where x is non-alphanumeric is a literal x
7666 \Q...\E treat enclosed characters as literal
7671 \a alarm, that is, the BEL character (hex 07)
7672 \cx "control-x", where x is any ASCII character
7674 \f form feed (hex 0C)
7676 \r carriage return (hex 0D)
7678 \0dd character with octal code 0dd
7679 \ddd character with octal code ddd, or backreference
7680 \o{ddd..} character with octal code ddd..
7681 \xhh character with hex code hh
7682 \x{hhh..} character with hex code hhh..
7684 Note that \0dd is always an octal code, and that \8 and \9 are the lit-
7685 eral characters "8" and "9".
7690 . any character except newline;
7691 in dotall mode, any character whatsoever
7692 \C one data unit, even in UTF mode (best avoided)
7694 \D a character that is not a decimal digit
7695 \h a horizontal white space character
7696 \H a character that is not a horizontal white space character
7697 \N a character that is not a newline
7698 \p{xx} a character with the xx property
7699 \P{xx} a character without the xx property
7700 \R a newline sequence
7701 \s a white space character
7702 \S a character that is not a white space character
7703 \v a vertical white space character
7704 \V a character that is not a vertical white space character
7705 \w a "word" character
7706 \W a "non-word" character
7707 \X a Unicode extended grapheme cluster
7709 By default, \d, \s, and \w match only ASCII characters, even in UTF-8
7710 mode or in the 16- bit and 32-bit libraries. However, if locale-spe-
7711 cific matching is happening, \s and \w may also match characters with
7712 code points in the range 128-255. If the PCRE_UCP option is set, the
7713 behaviour of these escape sequences is changed to use Unicode proper-
7714 ties and they match many more characters.
7717 GENERAL CATEGORY PROPERTIES FOR \p and \P
7727 Ll Lower case letter
7730 Lt Title case letter
7731 Lu Upper case letter
7745 Pc Connector punctuation
7747 Pe Close punctuation
7748 Pf Final punctuation
7749 Pi Initial punctuation
7750 Po Other punctuation
7756 Sm Mathematical symbol
7761 Zp Paragraph separator
7765 PCRE SPECIAL CATEGORY PROPERTIES FOR \p and \P
7767 Xan Alphanumeric: union of properties L and N
7768 Xps POSIX space: property Z or tab, NL, VT, FF, CR
7769 Xsp Perl space: property Z or tab, NL, VT, FF, CR
7770 Xuc Univerally-named character: one that can be
7771 represented by a Universal Character Name
7772 Xwd Perl word: property Xan or underscore
7774 Perl and POSIX space are now the same. Perl added VT to its space char-
7775 acter set at release 5.18 and PCRE changed at release 8.34.
7778 SCRIPT NAMES FOR \p AND \P
7780 Arabic, Armenian, Avestan, Balinese, Bamum, Batak, Bengali, Bopomofo,
7781 Brahmi, Braille, Buginese, Buhid, Canadian_Aboriginal, Carian, Chakma,
7782 Cham, Cherokee, Common, Coptic, Cuneiform, Cypriot, Cyrillic, Deseret,
7783 Devanagari, Egyptian_Hieroglyphs, Ethiopic, Georgian, Glagolitic,
7784 Gothic, Greek, Gujarati, Gurmukhi, Han, Hangul, Hanunoo, Hebrew, Hira-
7785 gana, Imperial_Aramaic, Inherited, Inscriptional_Pahlavi, Inscrip-
7786 tional_Parthian, Javanese, Kaithi, Kannada, Katakana, Kayah_Li,
7787 Kharoshthi, Khmer, Lao, Latin, Lepcha, Limbu, Linear_B, Lisu, Lycian,
7788 Lydian, Malayalam, Mandaic, Meetei_Mayek, Meroitic_Cursive,
7789 Meroitic_Hieroglyphs, Miao, Mongolian, Myanmar, New_Tai_Lue, Nko,
7790 Ogham, Old_Italic, Old_Persian, Old_South_Arabian, Old_Turkic,
7791 Ol_Chiki, Oriya, Osmanya, Phags_Pa, Phoenician, Rejang, Runic, Samari-
7792 tan, Saurashtra, Sharada, Shavian, Sinhala, Sora_Sompeng, Sundanese,
7793 Syloti_Nagri, Syriac, Tagalog, Tagbanwa, Tai_Le, Tai_Tham, Tai_Viet,
7794 Takri, Tamil, Telugu, Thaana, Thai, Tibetan, Tifinagh, Ugaritic, Vai,
7800 [...] positive character class
7801 [^...] negative character class
7802 [x-y] range (can be used for hex characters)
7803 [[:xxx:]] positive POSIX named set
7804 [[:^xxx:]] negative POSIX named set
7810 cntrl control character
7812 graph printing, excluding space
7813 lower lower case letter
7814 print printing, including space
7815 punct printing, excluding alphanumeric
7817 upper upper case letter
7819 xdigit hexadecimal digit
7821 In PCRE, POSIX character set names recognize only ASCII characters by
7822 default, but some of them use Unicode properties if PCRE_UCP is set.
7823 You can use \Q...\E inside a character class.
7829 ?+ 0 or 1, possessive
7832 *+ 0 or more, possessive
7835 ++ 1 or more, possessive
7838 {n,m} at least n, no more than m, greedy
7839 {n,m}+ at least n, no more than m, possessive
7840 {n,m}? at least n, no more than m, lazy
7841 {n,} n or more, greedy
7842 {n,}+ n or more, possessive
7843 {n,}? n or more, lazy
7846 ANCHORS AND SIMPLE ASSERTIONS
7849 \B not a word boundary
7851 also after internal newline in multiline mode
7854 also before newline at end of subject
7855 also before internal newline in multiline mode
7857 also before newline at end of subject
7859 \G first matching position in subject
7864 \K reset start of match
7866 \K is honoured in positive assertions, but ignored in negative ones.
7876 (...) capturing group
7877 (?<name>...) named capturing group (Perl)
7878 (?'name'...) named capturing group (Perl)
7879 (?P<name>...) named capturing group (Python)
7880 (?:...) non-capturing group
7881 (?|...) non-capturing group; reset group numbers for
7882 capturing groups in each alternative
7887 (?>...) atomic, non-capturing group
7892 (?#....) comment (not nestable)
7898 (?J) allow duplicate names
7900 (?s) single line (dotall)
7901 (?U) default ungreedy (lazy)
7902 (?x) extended (ignore white space)
7903 (?-...) unset option(s)
7905 The following are recognized only at the very start of a pattern or
7906 after one of the newline or \R options with similar syntax. More than
7907 one of them may appear.
7909 (*LIMIT_MATCH=d) set the match limit to d (decimal number)
7910 (*LIMIT_RECURSION=d) set the recursion limit to d (decimal number)
7911 (*NO_AUTO_POSSESS) no auto-possessification (PCRE_NO_AUTO_POSSESS)
7912 (*NO_START_OPT) no start-match optimization (PCRE_NO_START_OPTIMIZE)
7913 (*UTF8) set UTF-8 mode: 8-bit library (PCRE_UTF8)
7914 (*UTF16) set UTF-16 mode: 16-bit library (PCRE_UTF16)
7915 (*UTF32) set UTF-32 mode: 32-bit library (PCRE_UTF32)
7916 (*UTF) set appropriate UTF mode for the library in use
7917 (*UCP) set PCRE_UCP (use Unicode properties for \d etc)
7919 Note that LIMIT_MATCH and LIMIT_RECURSION can only reduce the value of
7920 the limits set by the caller of pcre_exec(), not increase them.
7925 These are recognized only at the very start of the pattern or after
7926 option settings with a similar syntax.
7928 (*CR) carriage return only
7930 (*CRLF) carriage return followed by linefeed
7931 (*ANYCRLF) all three of the above
7932 (*ANY) any Unicode newline sequence
7937 These are recognized only at the very start of the pattern or after
7938 option setting with a similar syntax.
7940 (*BSR_ANYCRLF) CR, LF, or CRLF
7941 (*BSR_UNICODE) any Unicode newline sequence
7944 LOOKAHEAD AND LOOKBEHIND ASSERTIONS
7946 (?=...) positive look ahead
7947 (?!...) negative look ahead
7948 (?<=...) positive look behind
7949 (?<!...) negative look behind
7951 Each top-level branch of a look behind must be of a fixed length.
7956 \n reference by number (can be ambiguous)
7957 \gn reference by number
7958 \g{n} reference by number
7959 \g{-n} relative reference by number
7960 \k<name> reference by name (Perl)
7961 \k'name' reference by name (Perl)
7962 \g{name} reference by name (Perl)
7963 \k{name} reference by name (.NET)
7964 (?P=name) reference by name (Python)
7967 SUBROUTINE REFERENCES (POSSIBLY RECURSIVE)
7969 (?R) recurse whole pattern
7970 (?n) call subpattern by absolute number
7971 (?+n) call subpattern by relative number
7972 (?-n) call subpattern by relative number
7973 (?&name) call subpattern by name (Perl)
7974 (?P>name) call subpattern by name (Python)
7975 \g<name> call subpattern by name (Oniguruma)
7976 \g'name' call subpattern by name (Oniguruma)
7977 \g<n> call subpattern by absolute number (Oniguruma)
7978 \g'n' call subpattern by absolute number (Oniguruma)
7979 \g<+n> call subpattern by relative number (PCRE extension)
7980 \g'+n' call subpattern by relative number (PCRE extension)
7981 \g<-n> call subpattern by relative number (PCRE extension)
7982 \g'-n' call subpattern by relative number (PCRE extension)
7985 CONDITIONAL PATTERNS
7987 (?(condition)yes-pattern)
7988 (?(condition)yes-pattern|no-pattern)
7990 (?(n)... absolute reference condition
7991 (?(+n)... relative reference condition
7992 (?(-n)... relative reference condition
7993 (?(<name>)... named reference condition (Perl)
7994 (?('name')... named reference condition (Perl)
7995 (?(name)... named reference condition (PCRE)
7996 (?(R)... overall recursion condition
7997 (?(Rn)... specific group recursion condition
7998 (?(R&name)... specific recursion condition
7999 (?(DEFINE)... define subpattern for reference
8000 (?(assert)... assertion condition
8003 BACKTRACKING CONTROL
8005 The following act immediately they are reached:
8007 (*ACCEPT) force successful match
8008 (*FAIL) force backtrack; synonym (*F)
8009 (*MARK:NAME) set name to be passed back; synonym (*:NAME)
8011 The following act only when a subsequent match failure causes a back-
8012 track to reach them. They all force a match failure, but they differ in
8013 what happens afterwards. Those that advance the start-of-match point do
8014 so only if the pattern is not anchored.
8016 (*COMMIT) overall failure, no advance of starting point
8017 (*PRUNE) advance to next starting character
8018 (*PRUNE:NAME) equivalent to (*MARK:NAME)(*PRUNE)
8019 (*SKIP) advance to current matching position
8020 (*SKIP:NAME) advance to position corresponding to an earlier
8021 (*MARK:NAME); if not found, the (*SKIP) is ignored
8022 (*THEN) local failure, backtrack to next alternation
8023 (*THEN:NAME) equivalent to (*MARK:NAME)(*THEN)
8029 (?Cn) callout with data n
8034 pcrepattern(3), pcreapi(3), pcrecallout(3), pcrematching(3), pcre(3).
8040 University Computing Service
8041 Cambridge CB2 3QH, England.
8046 Last updated: 08 January 2014
8047 Copyright (c) 1997-2014 University of Cambridge.
8048 ------------------------------------------------------------------------------
8051 PCREUNICODE(3) Library Functions Manual PCREUNICODE(3)
8056 PCRE - Perl-compatible regular expressions
8058 UTF-8, UTF-16, UTF-32, AND UNICODE PROPERTY SUPPORT
8060 As well as UTF-8 support, PCRE also supports UTF-16 (from release 8.30)
8061 and UTF-32 (from release 8.32), by means of two additional libraries.
8062 They can be built as well as, or instead of, the 8-bit library.
8067 In order process UTF-8 strings, you must build PCRE's 8-bit library
8068 with UTF support, and, in addition, you must call pcre_compile() with
8069 the PCRE_UTF8 option flag, or the pattern must start with the sequence
8070 (*UTF8) or (*UTF). When either of these is the case, both the pattern
8071 and any subject strings that are matched against it are treated as
8072 UTF-8 strings instead of strings of individual 1-byte characters.
8075 UTF-16 AND UTF-32 SUPPORT
8077 In order process UTF-16 or UTF-32 strings, you must build PCRE's 16-bit
8078 or 32-bit library with UTF support, and, in addition, you must call
8079 pcre16_compile() or pcre32_compile() with the PCRE_UTF16 or PCRE_UTF32
8080 option flag, as appropriate. Alternatively, the pattern must start with
8081 the sequence (*UTF16), (*UTF32), as appropriate, or (*UTF), which can
8082 be used with either library. When UTF mode is set, both the pattern and
8083 any subject strings that are matched against it are treated as UTF-16
8084 or UTF-32 strings instead of strings of individual 16-bit or 32-bit
8088 UTF SUPPORT OVERHEAD
8090 If you compile PCRE with UTF support, but do not use it at run time,
8091 the library will be a bit bigger, but the additional run time overhead
8092 is limited to testing the PCRE_UTF[8|16|32] flag occasionally, so
8093 should not be very big.
8096 UNICODE PROPERTY SUPPORT
8098 If PCRE is built with Unicode character property support (which implies
8099 UTF support), the escape sequences \p{..}, \P{..}, and \X can be used.
8100 The available properties that can be tested are limited to the general
8101 category properties such as Lu for an upper case letter or Nd for a
8102 decimal number, the Unicode script names such as Arabic or Han, and the
8103 derived properties Any and L&. Full lists is given in the pcrepattern
8104 and pcresyntax documentation. Only the short names for properties are
8105 supported. For example, \p{L} matches a letter. Its Perl synonym,
8106 \p{Letter}, is not supported. Furthermore, in Perl, many properties
8107 may optionally be prefixed by "Is", for compatibility with Perl 5.6.
8108 PCRE does not support this.
8110 Validity of UTF-8 strings
8112 When you set the PCRE_UTF8 flag, the byte strings passed as patterns
8113 and subjects are (by default) checked for validity on entry to the rel-
8114 evant functions. The entire string is checked before any other process-
8115 ing takes place. From release 7.3 of PCRE, the check is according the
8116 rules of RFC 3629, which are themselves derived from the Unicode speci-
8117 fication. Earlier releases of PCRE followed the rules of RFC 2279,
8118 which allows the full range of 31-bit values (0 to 0x7FFFFFFF). The
8119 current check allows only values in the range U+0 to U+10FFFF, exclud-
8120 ing the surrogate area. (From release 8.33 the so-called "non-charac-
8121 ter" code points are no longer excluded because Unicode corrigendum #9
8122 makes it clear that they should not be.)
8124 Characters in the "Surrogate Area" of Unicode are reserved for use by
8125 UTF-16, where they are used in pairs to encode codepoints with values
8126 greater than 0xFFFF. The code points that are encoded by UTF-16 pairs
8127 are available independently in the UTF-8 and UTF-32 encodings. (In
8128 other words, the whole surrogate thing is a fudge for UTF-16 which
8129 unfortunately messes up UTF-8 and UTF-32.)
8131 If an invalid UTF-8 string is passed to PCRE, an error return is given.
8132 At compile time, the only additional information is the offset to the
8133 first byte of the failing character. The run-time functions pcre_exec()
8134 and pcre_dfa_exec() also pass back this information, as well as a more
8135 detailed reason code if the caller has provided memory in which to do
8138 In some situations, you may already know that your strings are valid,
8139 and therefore want to skip these checks in order to improve perfor-
8140 mance, for example in the case of a long subject string that is being
8141 scanned repeatedly. If you set the PCRE_NO_UTF8_CHECK flag at compile
8142 time or at run time, PCRE assumes that the pattern or subject it is
8143 given (respectively) contains only valid UTF-8 codes. In this case, it
8144 does not diagnose an invalid UTF-8 string.
8146 Note that passing PCRE_NO_UTF8_CHECK to pcre_compile() just disables
8147 the check for the pattern; it does not also apply to subject strings.
8148 If you want to disable the check for a subject string you must pass
8149 this option to pcre_exec() or pcre_dfa_exec().
8151 If you pass an invalid UTF-8 string when PCRE_NO_UTF8_CHECK is set, the
8152 result is undefined and your program may crash.
8154 Validity of UTF-16 strings
8156 When you set the PCRE_UTF16 flag, the strings of 16-bit data units that
8157 are passed as patterns and subjects are (by default) checked for valid-
8158 ity on entry to the relevant functions. Values other than those in the
8159 surrogate range U+D800 to U+DFFF are independent code points. Values in
8160 the surrogate range must be used in pairs in the correct manner.
8162 If an invalid UTF-16 string is passed to PCRE, an error return is
8163 given. At compile time, the only additional information is the offset
8164 to the first data unit of the failing character. The run-time functions
8165 pcre16_exec() and pcre16_dfa_exec() also pass back this information, as
8166 well as a more detailed reason code if the caller has provided memory
8167 in which to do this.
8169 In some situations, you may already know that your strings are valid,
8170 and therefore want to skip these checks in order to improve perfor-
8171 mance. If you set the PCRE_NO_UTF16_CHECK flag at compile time or at
8172 run time, PCRE assumes that the pattern or subject it is given (respec-
8173 tively) contains only valid UTF-16 sequences. In this case, it does not
8174 diagnose an invalid UTF-16 string. However, if an invalid string is
8175 passed, the result is undefined.
8177 Validity of UTF-32 strings
8179 When you set the PCRE_UTF32 flag, the strings of 32-bit data units that
8180 are passed as patterns and subjects are (by default) checked for valid-
8181 ity on entry to the relevant functions. This check allows only values
8182 in the range U+0 to U+10FFFF, excluding the surrogate area U+D800 to
8185 If an invalid UTF-32 string is passed to PCRE, an error return is
8186 given. At compile time, the only additional information is the offset
8187 to the first data unit of the failing character. The run-time functions
8188 pcre32_exec() and pcre32_dfa_exec() also pass back this information, as
8189 well as a more detailed reason code if the caller has provided memory
8190 in which to do this.
8192 In some situations, you may already know that your strings are valid,
8193 and therefore want to skip these checks in order to improve perfor-
8194 mance. If you set the PCRE_NO_UTF32_CHECK flag at compile time or at
8195 run time, PCRE assumes that the pattern or subject it is given (respec-
8196 tively) contains only valid UTF-32 sequences. In this case, it does not
8197 diagnose an invalid UTF-32 string. However, if an invalid string is
8198 passed, the result is undefined.
8200 General comments about UTF modes
8202 1. Codepoints less than 256 can be specified in patterns by either
8203 braced or unbraced hexadecimal escape sequences (for example, \x{b3} or
8204 \xb3). Larger values have to use braced sequences.
8206 2. Octal numbers up to \777 are recognized, and in UTF-8 mode they
8207 match two-byte characters for values greater than \177.
8209 3. Repeat quantifiers apply to complete UTF characters, not to individ-
8210 ual data units, for example: \x{100}{3}.
8212 4. The dot metacharacter matches one UTF character instead of a single
8215 5. The escape sequence \C can be used to match a single byte in UTF-8
8216 mode, or a single 16-bit data unit in UTF-16 mode, or a single 32-bit
8217 data unit in UTF-32 mode, but its use can lead to some strange effects
8218 because it breaks up multi-unit characters (see the description of \C
8219 in the pcrepattern documentation). The use of \C is not supported in
8220 the alternative matching function pcre[16|32]_dfa_exec(), nor is it
8221 supported in UTF mode by the JIT optimization of pcre[16|32]_exec(). If
8222 JIT optimization is requested for a UTF pattern that contains \C, it
8223 will not succeed, and so the matching will be carried out by the normal
8224 interpretive function.
8226 6. The character escapes \b, \B, \d, \D, \s, \S, \w, and \W correctly
8227 test characters of any code value, but, by default, the characters that
8228 PCRE recognizes as digits, spaces, or word characters remain the same
8229 set as in non-UTF mode, all with values less than 256. This remains
8230 true even when PCRE is built to include Unicode property support,
8231 because to do otherwise would slow down PCRE in many common cases. Note
8232 in particular that this applies to \b and \B, because they are defined
8233 in terms of \w and \W. If you really want to test for a wider sense of,
8234 say, "digit", you can use explicit Unicode property tests such as
8235 \p{Nd}. Alternatively, if you set the PCRE_UCP option, the way that the
8236 character escapes work is changed so that Unicode properties are used
8237 to determine which characters match. There are more details in the sec-
8238 tion on generic character types in the pcrepattern documentation.
8240 7. Similarly, characters that match the POSIX named character classes
8241 are all low-valued characters, unless the PCRE_UCP option is set.
8243 8. However, the horizontal and vertical white space matching escapes
8244 (\h, \H, \v, and \V) do match all the appropriate Unicode characters,
8245 whether or not PCRE_UCP is set.
8247 9. Case-insensitive matching applies only to characters whose values
8248 are less than 128, unless PCRE is built with Unicode property support.
8249 A few Unicode characters such as Greek sigma have more than two code-
8250 points that are case-equivalent. Up to and including PCRE release 8.31,
8251 only one-to-one case mappings were supported, but later releases (with
8252 Unicode property support) do treat as case-equivalent all versions of
8253 characters such as Greek sigma.
8259 University Computing Service
8260 Cambridge CB2 3QH, England.
8265 Last updated: 27 February 2013
8266 Copyright (c) 1997-2013 University of Cambridge.
8267 ------------------------------------------------------------------------------
8270 PCREJIT(3) Library Functions Manual PCREJIT(3)
8275 PCRE - Perl-compatible regular expressions
8277 PCRE JUST-IN-TIME COMPILER SUPPORT
8279 Just-in-time compiling is a heavyweight optimization that can greatly
8280 speed up pattern matching. However, it comes at the cost of extra pro-
8281 cessing before the match is performed. Therefore, it is of most benefit
8282 when the same pattern is going to be matched many times. This does not
8283 necessarily mean many calls of a matching function; if the pattern is
8284 not anchored, matching attempts may take place many times at various
8285 positions in the subject, even for a single call. Therefore, if the
8286 subject string is very long, it may still pay to use JIT for one-off
8289 JIT support applies only to the traditional Perl-compatible matching
8290 function. It does not apply when the DFA matching function is being
8291 used. The code for this support was written by Zoltan Herczeg.
8294 8-BIT, 16-BIT AND 32-BIT SUPPORT
8296 JIT support is available for all of the 8-bit, 16-bit and 32-bit PCRE
8297 libraries. To keep this documentation simple, only the 8-bit interface
8298 is described in what follows. If you are using the 16-bit library, sub-
8299 stitute the 16-bit functions and 16-bit structures (for example,
8300 pcre16_jit_stack instead of pcre_jit_stack). If you are using the
8301 32-bit library, substitute the 32-bit functions and 32-bit structures
8302 (for example, pcre32_jit_stack instead of pcre_jit_stack).
8305 AVAILABILITY OF JIT SUPPORT
8307 JIT support is an optional feature of PCRE. The "configure" option
8308 --enable-jit (or equivalent CMake option) must be set when PCRE is
8309 built if you want to use JIT. The support is limited to the following
8312 ARM v5, v7, and Thumb2
8313 Intel x86 32-bit and 64-bit
8315 Power PC 32-bit and 64-bit
8316 SPARC 32-bit (experimental)
8318 If --enable-jit is set on an unsupported platform, compilation fails.
8320 A program that is linked with PCRE 8.20 or later can tell if JIT sup-
8321 port is available by calling pcre_config() with the PCRE_CONFIG_JIT
8322 option. The result is 1 when JIT is available, and 0 otherwise. How-
8323 ever, a simple program does not need to check this in order to use JIT.
8324 The normal API is implemented in a way that falls back to the interpre-
8325 tive code if JIT is not available. For programs that need the best pos-
8326 sible performance, there is also a "fast path" API that is JIT-spe-
8329 If your program may sometimes be linked with versions of PCRE that are
8330 older than 8.20, but you want to use JIT when it is available, you can
8331 test the values of PCRE_MAJOR and PCRE_MINOR, or the existence of a JIT
8332 macro such as PCRE_CONFIG_JIT, for compile-time control of your code.
8337 You have to do two things to make use of the JIT support in the sim-
8340 (1) Call pcre_study() with the PCRE_STUDY_JIT_COMPILE option for
8341 each compiled pattern, and pass the resulting pcre_extra block to
8344 (2) Use pcre_free_study() to free the pcre_extra block when it is
8345 no longer needed, instead of just freeing it yourself. This
8347 any JIT data is also freed.
8349 For a program that may be linked with pre-8.20 versions of PCRE, you
8352 #ifndef PCRE_STUDY_JIT_COMPILE
8353 #define PCRE_STUDY_JIT_COMPILE 0
8356 so that no option is passed to pcre_study(), and then use something
8357 like this to free the study data:
8359 #ifdef PCRE_CONFIG_JIT
8360 pcre_free_study(study_ptr);
8362 pcre_free(study_ptr);
8365 PCRE_STUDY_JIT_COMPILE requests the JIT compiler to generate code for
8366 complete matches. If you want to run partial matches using the
8367 PCRE_PARTIAL_HARD or PCRE_PARTIAL_SOFT options of pcre_exec(), you
8368 should set one or both of the following options in addition to, or
8369 instead of, PCRE_STUDY_JIT_COMPILE when you call pcre_study():
8371 PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE
8372 PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE
8374 The JIT compiler generates different optimized code for each of the
8375 three modes (normal, soft partial, hard partial). When pcre_exec() is
8376 called, the appropriate code is run if it is available. Otherwise, the
8377 pattern is matched using interpretive code.
8379 In some circumstances you may need to call additional functions. These
8380 are described in the section entitled "Controlling the JIT stack"
8383 If JIT support is not available, PCRE_STUDY_JIT_COMPILE etc. are
8384 ignored, and no JIT data is created. Otherwise, the compiled pattern is
8385 passed to the JIT compiler, which turns it into machine code that exe-
8386 cutes much faster than the normal interpretive code. When pcre_exec()
8387 is passed a pcre_extra block containing a pointer to JIT code of the
8388 appropriate mode (normal or hard/soft partial), it obeys that code
8389 instead of running the interpreter. The result is identical, but the
8390 compiled JIT code runs much faster.
8392 There are some pcre_exec() options that are not supported for JIT exe-
8393 cution. There are also some pattern items that JIT cannot handle.
8394 Details are given below. In both cases, execution automatically falls
8395 back to the interpretive code. If you want to know whether JIT was
8396 actually used for a particular match, you should arrange for a JIT
8397 callback function to be set up as described in the section entitled
8398 "Controlling the JIT stack" below, even if you do not need to supply a
8399 non-default JIT stack. Such a callback function is called whenever JIT
8400 code is about to be obeyed. If the execution options are not right for
8401 JIT execution, the callback function is not obeyed.
8403 If the JIT compiler finds an unsupported item, no JIT data is gener-
8404 ated. You can find out if JIT execution is available after studying a
8405 pattern by calling pcre_fullinfo() with the PCRE_INFO_JIT option. A
8406 result of 1 means that JIT compilation was successful. A result of 0
8407 means that JIT support is not available, or the pattern was not studied
8408 with PCRE_STUDY_JIT_COMPILE etc., or the JIT compiler was not able to
8411 Once a pattern has been studied, with or without JIT, it can be used as
8412 many times as you like for matching different subject strings.
8415 UNSUPPORTED OPTIONS AND PATTERN ITEMS
8417 The only pcre_exec() options that are supported for JIT execution are
8418 PCRE_NO_UTF8_CHECK, PCRE_NO_UTF16_CHECK, PCRE_NO_UTF32_CHECK, PCRE_NOT-
8419 BOL, PCRE_NOTEOL, PCRE_NOTEMPTY, PCRE_NOTEMPTY_ATSTART, PCRE_PAR-
8420 TIAL_HARD, and PCRE_PARTIAL_SOFT.
8422 The only unsupported pattern items are \C (match a single data unit)
8423 when running in a UTF mode, and a callout immediately before an asser-
8424 tion condition in a conditional group.
8427 RETURN VALUES FROM JIT EXECUTION
8429 When a pattern is matched using JIT execution, the return values are
8430 the same as those given by the interpretive pcre_exec() code, with the
8431 addition of one new error code: PCRE_ERROR_JIT_STACKLIMIT. This means
8432 that the memory used for the JIT stack was insufficient. See "Control-
8433 ling the JIT stack" below for a discussion of JIT stack usage. For com-
8434 patibility with the interpretive pcre_exec() code, no more than two-
8435 thirds of the ovector argument is used for passing back captured sub-
8438 The error code PCRE_ERROR_MATCHLIMIT is returned by the JIT code if
8439 searching a very large pattern tree goes on for too long, as it is in
8440 the same circumstance when JIT is not used, but the details of exactly
8441 what is counted are not the same. The PCRE_ERROR_RECURSIONLIMIT error
8442 code is never returned by JIT execution.
8445 SAVING AND RESTORING COMPILED PATTERNS
8447 The code that is generated by the JIT compiler is architecture-spe-
8448 cific, and is also position dependent. For those reasons it cannot be
8449 saved (in a file or database) and restored later like the bytecode and
8450 other data of a compiled pattern. Saving and restoring compiled pat-
8451 terns is not something many people do. More detail about this facility
8452 is given in the pcreprecompile documentation. It should be possible to
8453 run pcre_study() on a saved and restored pattern, and thereby recreate
8454 the JIT data, but because JIT compilation uses significant resources,
8455 it is probably not worth doing this; you might as well recompile the
8459 CONTROLLING THE JIT STACK
8461 When the compiled JIT code runs, it needs a block of memory to use as a
8462 stack. By default, it uses 32K on the machine stack. However, some
8463 large or complicated patterns need more than this. The error
8464 PCRE_ERROR_JIT_STACKLIMIT is given when there is not enough stack.
8465 Three functions are provided for managing blocks of memory for use as
8466 JIT stacks. There is further discussion about the use of JIT stacks in
8467 the section entitled "JIT stack FAQ" below.
8469 The pcre_jit_stack_alloc() function creates a JIT stack. Its arguments
8470 are a starting size and a maximum size, and it returns a pointer to an
8471 opaque structure of type pcre_jit_stack, or NULL if there is an error.
8472 The pcre_jit_stack_free() function can be used to free a stack that is
8473 no longer needed. (For the technically minded: the address space is
8474 allocated by mmap or VirtualAlloc.)
8476 JIT uses far less memory for recursion than the interpretive code, and
8477 a maximum stack size of 512K to 1M should be more than enough for any
8480 The pcre_assign_jit_stack() function specifies which stack JIT code
8481 should use. Its arguments are as follows:
8484 pcre_jit_callback callback
8487 The extra argument must be the result of studying a pattern with
8488 PCRE_STUDY_JIT_COMPILE etc. There are three cases for the values of the
8491 (1) If callback is NULL and data is NULL, an internal 32K block
8492 on the machine stack is used.
8494 (2) If callback is NULL and data is not NULL, data must be
8495 a valid JIT stack, the result of calling pcre_jit_stack_alloc().
8497 (3) If callback is not NULL, it must point to a function that is
8498 called with data as an argument at the start of matching, in
8499 order to set up a JIT stack. If the return from the callback
8500 function is NULL, the internal 32K stack is used; otherwise the
8501 return value must be a valid JIT stack, the result of calling
8502 pcre_jit_stack_alloc().
8504 A callback function is obeyed whenever JIT code is about to be run; it
8505 is not obeyed when pcre_exec() is called with options that are incom-
8506 patible for JIT execution. A callback function can therefore be used to
8507 determine whether a match operation was executed by JIT or by the
8510 You may safely use the same JIT stack for more than one pattern (either
8511 by assigning directly or by callback), as long as the patterns are all
8512 matched sequentially in the same thread. In a multithread application,
8513 if you do not specify a JIT stack, or if you assign or pass back NULL
8514 from a callback, that is thread-safe, because each thread has its own
8515 machine stack. However, if you assign or pass back a non-NULL JIT
8516 stack, this must be a different stack for each thread so that the
8517 application is thread-safe.
8519 Strictly speaking, even more is allowed. You can assign the same non-
8520 NULL stack to any number of patterns as long as they are not used for
8521 matching by multiple threads at the same time. For example, you can
8522 assign the same stack to all compiled patterns, and use a global mutex
8523 in the callback to wait until the stack is available for use. However,
8524 this is an inefficient solution, and not recommended.
8526 This is a suggestion for how a multithreaded program that needs to set
8527 up non-default JIT stacks might operate:
8529 During thread initalization
8530 thread_local_var = pcre_jit_stack_alloc(...)
8533 pcre_jit_stack_free(thread_local_var)
8535 Use a one-line callback function
8536 return thread_local_var
8538 All the functions described in this section do nothing if JIT is not
8539 available, and pcre_assign_jit_stack() does nothing unless the extra
8540 argument is non-NULL and points to a pcre_extra block that is the
8541 result of a successful study with PCRE_STUDY_JIT_COMPILE etc.
8546 (1) Why do we need JIT stacks?
8548 PCRE (and JIT) is a recursive, depth-first engine, so it needs a stack
8549 where the local data of the current node is pushed before checking its
8550 child nodes. Allocating real machine stack on some platforms is diffi-
8551 cult. For example, the stack chain needs to be updated every time if we
8552 extend the stack on PowerPC. Although it is possible, its updating
8553 time overhead decreases performance. So we do the recursion in memory.
8555 (2) Why don't we simply allocate blocks of memory with malloc()?
8557 Modern operating systems have a nice feature: they can reserve an
8558 address space instead of allocating memory. We can safely allocate mem-
8559 ory pages inside this address space, so the stack could grow without
8560 moving memory data (this is important because of pointers). Thus we can
8561 allocate 1M address space, and use only a single memory page (usually
8562 4K) if that is enough. However, we can still grow up to 1M anytime if
8565 (3) Who "owns" a JIT stack?
8567 The owner of the stack is the user program, not the JIT studied pattern
8568 or anything else. The user program must ensure that if a stack is used
8569 by pcre_exec(), (that is, it is assigned to the pattern currently run-
8570 ning), that stack must not be used by any other threads (to avoid over-
8571 writing the same memory area). The best practice for multithreaded pro-
8572 grams is to allocate a stack for each thread, and return this stack
8573 through the JIT callback function.
8575 (4) When should a JIT stack be freed?
8577 You can free a JIT stack at any time, as long as it will not be used by
8578 pcre_exec() again. When you assign the stack to a pattern, only a
8579 pointer is set. There is no reference counting or any other magic. You
8580 can free the patterns and stacks in any order, anytime. Just do not
8581 call pcre_exec() with a pattern pointing to an already freed stack, as
8582 that will cause SEGFAULT. (Also, do not free a stack currently used by
8583 pcre_exec() in another thread). You can also replace the stack for a
8584 pattern at any time. You can even free the previous stack before
8585 assigning a replacement.
8587 (5) Should I allocate/free a stack every time before/after calling
8590 No, because this is too costly in terms of resources. However, you
8591 could implement some clever idea which release the stack if it is not
8592 used in let's say two minutes. The JIT callback can help to achieve
8593 this without keeping a list of the currently JIT studied patterns.
8595 (6) OK, the stack is for long term memory allocation. But what happens
8596 if a pattern causes stack overflow with a stack of 1M? Is that 1M kept
8597 until the stack is freed?
8599 Especially on embedded sytems, it might be a good idea to release mem-
8600 ory sometimes without freeing the stack. There is no API for this at
8601 the moment. Probably a function call which returns with the currently
8602 allocated memory for any stack and another which allows releasing mem-
8603 ory (shrinking the stack) would be a good idea if someone needs this.
8605 (7) This is too much of a headache. Isn't there any better solution for
8608 No, thanks to Windows. If POSIX threads were used everywhere, we could
8609 throw out this complicated API.
8614 This is a single-threaded example that specifies a JIT stack without
8621 pcre_jit_stack *jit_stack;
8623 re = pcre_compile(pattern, 0, &error, &erroffset, NULL);
8624 /* Check for errors */
8625 extra = pcre_study(re, PCRE_STUDY_JIT_COMPILE, &error);
8626 jit_stack = pcre_jit_stack_alloc(32*1024, 512*1024);
8627 /* Check for error (NULL) */
8628 pcre_assign_jit_stack(extra, NULL, jit_stack);
8629 rc = pcre_exec(re, extra, subject, length, 0, 0, ovector, 30);
8632 pcre_free_study(extra);
8633 pcre_jit_stack_free(jit_stack);
8638 Because the API described above falls back to interpreted execution
8639 when JIT is not available, it is convenient for programs that are writ-
8640 ten for general use in many environments. However, calling JIT via
8641 pcre_exec() does have a performance impact. Programs that are written
8642 for use where JIT is known to be available, and which need the best
8643 possible performance, can instead use a "fast path" API to call JIT
8644 execution directly instead of calling pcre_exec() (obviously only for
8645 patterns that have been successfully studied by JIT).
8647 The fast path function is called pcre_jit_exec(), and it takes exactly
8648 the same arguments as pcre_exec(), plus one additional argument that
8649 must point to a JIT stack. The JIT stack arrangements described above
8650 do not apply. The return values are the same as for pcre_exec().
8652 When you call pcre_exec(), as well as testing for invalid options, a
8653 number of other sanity checks are performed on the arguments. For exam-
8654 ple, if the subject pointer is NULL, or its length is negative, an
8655 immediate error is given. Also, unless PCRE_NO_UTF[8|16|32] is set, a
8656 UTF subject string is tested for validity. In the interests of speed,
8657 these checks do not happen on the JIT fast path, and if invalid data is
8658 passed, the result is undefined.
8660 Bypassing the sanity checks and the pcre_exec() wrapping can give
8661 speedups of more than 10%.
8671 Philip Hazel (FAQ by Zoltan Herczeg)
8672 University Computing Service
8673 Cambridge CB2 3QH, England.
8678 Last updated: 17 March 2013
8679 Copyright (c) 1997-2013 University of Cambridge.
8680 ------------------------------------------------------------------------------
8683 PCREPARTIAL(3) Library Functions Manual PCREPARTIAL(3)
8688 PCRE - Perl-compatible regular expressions
8690 PARTIAL MATCHING IN PCRE
8692 In normal use of PCRE, if the subject string that is passed to a match-
8693 ing function matches as far as it goes, but is too short to match the
8694 entire pattern, PCRE_ERROR_NOMATCH is returned. There are circumstances
8695 where it might be helpful to distinguish this case from other cases in
8696 which there is no match.
8698 Consider, for example, an application where a human is required to type
8699 in data for a field with specific formatting requirements. An example
8700 might be a date in the form ddmmmyy, defined by this pattern:
8702 ^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$
8704 If the application sees the user's keystrokes one by one, and can check
8705 that what has been typed so far is potentially valid, it is able to
8706 raise an error as soon as a mistake is made, by beeping and not
8707 reflecting the character that has been typed, for example. This immedi-
8708 ate feedback is likely to be a better user interface than a check that
8709 is delayed until the entire string has been entered. Partial matching
8710 can also be useful when the subject string is very long and is not all
8713 PCRE supports partial matching by means of the PCRE_PARTIAL_SOFT and
8714 PCRE_PARTIAL_HARD options, which can be set when calling any of the
8715 matching functions. For backwards compatibility, PCRE_PARTIAL is a syn-
8716 onym for PCRE_PARTIAL_SOFT. The essential difference between the two
8717 options is whether or not a partial match is preferred to an alterna-
8718 tive complete match, though the details differ between the two types of
8719 matching function. If both options are set, PCRE_PARTIAL_HARD takes
8722 If you want to use partial matching with just-in-time optimized code,
8723 you must call pcre_study(), pcre16_study() or pcre32_study() with one
8724 or both of these options:
8726 PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE
8727 PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE
8729 PCRE_STUDY_JIT_COMPILE should also be set if you are going to run non-
8730 partial matches on the same pattern. If the appropriate JIT study mode
8731 has not been set for a match, the interpretive matching code is used.
8733 Setting a partial matching option disables two of PCRE's standard opti-
8734 mizations. PCRE remembers the last literal data unit in a pattern, and
8735 abandons matching immediately if it is not present in the subject
8736 string. This optimization cannot be used for a subject string that
8737 might match only partially. If the pattern was studied, PCRE knows the
8738 minimum length of a matching string, and does not bother to run the
8739 matching function on shorter strings. This optimization is also dis-
8740 abled for partial matching.
8743 PARTIAL MATCHING USING pcre_exec() OR pcre[16|32]_exec()
8745 A partial match occurs during a call to pcre_exec() or
8746 pcre[16|32]_exec() when the end of the subject string is reached suc-
8747 cessfully, but matching cannot continue because more characters are
8748 needed. However, at least one character in the subject must have been
8749 inspected. This character need not form part of the final matched
8750 string; lookbehind assertions and the \K escape sequence provide ways
8751 of inspecting characters before the start of a matched substring. The
8752 requirement for inspecting at least one character exists because an
8753 empty string can always be matched; without such a restriction there
8754 would always be a partial match of an empty string at the end of the
8757 If there are at least two slots in the offsets vector when a partial
8758 match is returned, the first slot is set to the offset of the earliest
8759 character that was inspected. For convenience, the second offset points
8760 to the end of the subject so that a substring can easily be identified.
8761 If there are at least three slots in the offsets vector, the third slot
8762 is set to the offset of the character where matching started.
8764 For the majority of patterns, the contents of the first and third slots
8765 will be the same. However, for patterns that contain lookbehind asser-
8766 tions, or begin with \b or \B, characters before the one where matching
8767 started may have been inspected while carrying out the match. For exam-
8768 ple, consider this pattern:
8772 This pattern matches "123", but only if it is preceded by "abc". If the
8773 subject string is "xyzabc12", the first two offsets after a partial
8774 match are for the substring "abc12", because all these characters were
8775 inspected. However, the third offset is set to 6, because that is the
8776 offset where matching began.
8778 What happens when a partial match is identified depends on which of the
8779 two partial matching options are set.
8781 PCRE_PARTIAL_SOFT WITH pcre_exec() OR pcre[16|32]_exec()
8783 If PCRE_PARTIAL_SOFT is set when pcre_exec() or pcre[16|32]_exec()
8784 identifies a partial match, the partial match is remembered, but match-
8785 ing continues as normal, and other alternatives in the pattern are
8786 tried. If no complete match can be found, PCRE_ERROR_PARTIAL is
8787 returned instead of PCRE_ERROR_NOMATCH.
8789 This option is "soft" because it prefers a complete match over a par-
8790 tial match. All the various matching items in a pattern behave as if
8791 the subject string is potentially complete. For example, \z, \Z, and $
8792 match at the end of the subject, as normal, and for \b and \B the end
8793 of the subject is treated as a non-alphanumeric.
8795 If there is more than one partial match, the first one that was found
8796 provides the data that is returned. Consider this pattern:
8800 If this is matched against the subject string "abc123dog", both alter-
8801 natives fail to match, but the end of the subject is reached during
8802 matching, so PCRE_ERROR_PARTIAL is returned. The offsets are set to 3
8803 and 9, identifying "123dog" as the first partial match that was found.
8804 (In this example, there are two partial matches, because "dog" on its
8805 own partially matches the second alternative.)
8807 PCRE_PARTIAL_HARD WITH pcre_exec() OR pcre[16|32]_exec()
8809 If PCRE_PARTIAL_HARD is set for pcre_exec() or pcre[16|32]_exec(),
8810 PCRE_ERROR_PARTIAL is returned as soon as a partial match is found,
8811 without continuing to search for possible complete matches. This option
8812 is "hard" because it prefers an earlier partial match over a later com-
8813 plete match. For this reason, the assumption is made that the end of
8814 the supplied subject string may not be the true end of the available
8815 data, and so, if \z, \Z, \b, \B, or $ are encountered at the end of the
8816 subject, the result is PCRE_ERROR_PARTIAL, provided that at least one
8817 character in the subject has been inspected.
8819 Setting PCRE_PARTIAL_HARD also affects the way UTF-8 and UTF-16 subject
8820 strings are checked for validity. Normally, an invalid sequence causes
8821 the error PCRE_ERROR_BADUTF8 or PCRE_ERROR_BADUTF16. However, in the
8822 special case of a truncated character at the end of the subject,
8823 PCRE_ERROR_SHORTUTF8 or PCRE_ERROR_SHORTUTF16 is returned when
8824 PCRE_PARTIAL_HARD is set.
8826 Comparing hard and soft partial matching
8828 The difference between the two partial matching options can be illus-
8829 trated by a pattern such as:
8833 This matches either "dog" or "dogsbody", greedily (that is, it prefers
8834 the longer string if possible). If it is matched against the string
8835 "dog" with PCRE_PARTIAL_SOFT, it yields a complete match for "dog".
8836 However, if PCRE_PARTIAL_HARD is set, the result is PCRE_ERROR_PARTIAL.
8837 On the other hand, if the pattern is made ungreedy the result is dif-
8842 In this case the result is always a complete match because that is
8843 found first, and matching never continues after finding a complete
8844 match. It might be easier to follow this explanation by thinking of the
8845 two patterns like this:
8847 /dog(sbody)?/ is the same as /dogsbody|dog/
8848 /dog(sbody)??/ is the same as /dog|dogsbody/
8850 The second pattern will never match "dogsbody", because it will always
8851 find the shorter match first.
8854 PARTIAL MATCHING USING pcre_dfa_exec() OR pcre[16|32]_dfa_exec()
8856 The DFA functions move along the subject string character by character,
8857 without backtracking, searching for all possible matches simultane-
8858 ously. If the end of the subject is reached before the end of the pat-
8859 tern, there is the possibility of a partial match, again provided that
8860 at least one character has been inspected.
8862 When PCRE_PARTIAL_SOFT is set, PCRE_ERROR_PARTIAL is returned only if
8863 there have been no complete matches. Otherwise, the complete matches
8864 are returned. However, if PCRE_PARTIAL_HARD is set, a partial match
8865 takes precedence over any complete matches. The portion of the string
8866 that was inspected when the longest partial match was found is set as
8867 the first matching string, provided there are at least two slots in the
8870 Because the DFA functions always search for all possible matches, and
8871 there is no difference between greedy and ungreedy repetition, their
8872 behaviour is different from the standard functions when PCRE_PAR-
8873 TIAL_HARD is set. Consider the string "dog" matched against the
8874 ungreedy pattern shown above:
8878 Whereas the standard functions stop as soon as they find the complete
8879 match for "dog", the DFA functions also find the partial match for
8880 "dogsbody", and so return that when PCRE_PARTIAL_HARD is set.
8883 PARTIAL MATCHING AND WORD BOUNDARIES
8885 If a pattern ends with one of sequences \b or \B, which test for word
8886 boundaries, partial matching with PCRE_PARTIAL_SOFT can give counter-
8887 intuitive results. Consider this pattern:
8891 This matches "cat", provided there is a word boundary at either end. If
8892 the subject string is "the cat", the comparison of the final "t" with a
8893 following character cannot take place, so a partial match is found.
8894 However, normal matching carries on, and \b matches at the end of the
8895 subject when the last character is a letter, so a complete match is
8896 found. The result, therefore, is not PCRE_ERROR_PARTIAL. Using
8897 PCRE_PARTIAL_HARD in this case does yield PCRE_ERROR_PARTIAL, because
8898 then the partial match takes precedence.
8901 FORMERLY RESTRICTED PATTERNS
8903 For releases of PCRE prior to 8.00, because of the way certain internal
8904 optimizations were implemented in the pcre_exec() function, the
8905 PCRE_PARTIAL option (predecessor of PCRE_PARTIAL_SOFT) could not be
8906 used with all patterns. From release 8.00 onwards, the restrictions no
8907 longer apply, and partial matching with can be requested for any pat-
8910 Items that were formerly restricted were repeated single characters and
8911 repeated metasequences. If PCRE_PARTIAL was set for a pattern that did
8912 not conform to the restrictions, pcre_exec() returned the error code
8913 PCRE_ERROR_BADPARTIAL (-13). This error code is no longer in use. The
8914 PCRE_INFO_OKPARTIAL call to pcre_fullinfo() to find out if a compiled
8915 pattern can be used for partial matching now always returns 1.
8918 EXAMPLE OF PARTIAL MATCHING USING PCRETEST
8920 If the escape sequence \P is present in a pcretest data line, the
8921 PCRE_PARTIAL_SOFT option is used for the match. Here is a run of
8922 pcretest that uses the date example quoted above:
8924 re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/
8929 Partial match: 23dec3
8937 The first data string is matched completely, so pcretest shows the
8938 matched substrings. The remaining four strings do not match the com-
8939 plete pattern, but the first two are partial matches. Similar output is
8940 obtained if DFA matching is used.
8942 If the escape sequence \P is present more than once in a pcretest data
8943 line, the PCRE_PARTIAL_HARD option is set for the match.
8946 MULTI-SEGMENT MATCHING WITH pcre_dfa_exec() OR pcre[16|32]_dfa_exec()
8948 When a partial match has been found using a DFA matching function, it
8949 is possible to continue the match by providing additional subject data
8950 and calling the function again with the same compiled regular expres-
8951 sion, this time setting the PCRE_DFA_RESTART option. You must pass the
8952 same working space as before, because this is where details of the pre-
8953 vious partial match are stored. Here is an example using pcretest,
8954 using the \R escape sequence to set the PCRE_DFA_RESTART option (\D
8955 specifies the use of the DFA matching function):
8957 re> /^\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d$/
8963 The first call has "23ja" as the subject, and requests partial match-
8964 ing; the second call has "n05" as the subject for the continued
8965 (restarted) match. Notice that when the match is complete, only the
8966 last part is shown; PCRE does not retain the previously partially-
8967 matched string. It is up to the calling program to do that if it needs
8970 That means that, for an unanchored pattern, if a continued match fails,
8971 it is not possible to try again at a new starting point. All this
8972 facility is capable of doing is continuing with the previous match
8973 attempt. In the previous example, if the second set of data is "ug23"
8974 the result is no match, even though there would be a match for "aug23"
8975 if the entire string were given at once. Depending on the application,
8976 this may or may not be what you want. The only way to allow for start-
8977 ing again at the next character is to retain the matched part of the
8978 subject and try a new complete match.
8980 You can set the PCRE_PARTIAL_SOFT or PCRE_PARTIAL_HARD options with
8981 PCRE_DFA_RESTART to continue partial matching over multiple segments.
8982 This facility can be used to pass very long subject strings to the DFA
8986 MULTI-SEGMENT MATCHING WITH pcre_exec() OR pcre[16|32]_exec()
8988 From release 8.00, the standard matching functions can also be used to
8989 do multi-segment matching. Unlike the DFA functions, it is not possible
8990 to restart the previous match with a new segment of data. Instead, new
8991 data must be added to the previous subject string, and the entire match
8992 re-run, starting from the point where the partial match occurred. Ear-
8993 lier data can be discarded.
8995 It is best to use PCRE_PARTIAL_HARD in this situation, because it does
8996 not treat the end of a segment as the end of the subject when matching
8997 \z, \Z, \b, \B, and $. Consider an unanchored pattern that matches
9000 re> /\d?\d(jan|feb|mar|apr|may|jun|jul|aug|sep|oct|nov|dec)\d\d/
9001 data> The date is 23ja\P\P
9004 At this stage, an application could discard the text preceding "23ja",
9005 add on text from the next segment, and call the matching function
9006 again. Unlike the DFA matching functions, the entire matching string
9007 must always be available, and the complete matching process occurs for
9008 each call, so more memory and more processing time is needed.
9010 Note: If the pattern contains lookbehind assertions, or \K, or starts
9011 with \b or \B, the string that is returned for a partial match includes
9012 characters that precede the start of what would be returned for a com-
9013 plete match, because it contains all the characters that were inspected
9014 during the partial match.
9017 ISSUES WITH MULTI-SEGMENT MATCHING
9019 Certain types of pattern may give problems with multi-segment matching,
9020 whichever matching function is used.
9022 1. If the pattern contains a test for the beginning of a line, you need
9023 to pass the PCRE_NOTBOL option when the subject string for any call
9024 does start at the beginning of a line. There is also a PCRE_NOTEOL
9025 option, but in practice when doing multi-segment matching you should be
9026 using PCRE_PARTIAL_HARD, which includes the effect of PCRE_NOTEOL.
9028 2. Lookbehind assertions that have already been obeyed are catered for
9029 in the offsets that are returned for a partial match. However a lookbe-
9030 hind assertion later in the pattern could require even earlier charac-
9031 ters to be inspected. You can handle this case by using the
9032 PCRE_INFO_MAXLOOKBEHIND option of the pcre_fullinfo() or
9033 pcre[16|32]_fullinfo() functions to obtain the length of the longest
9034 lookbehind in the pattern. This length is given in characters, not
9035 bytes. If you always retain at least that many characters before the
9036 partially matched string, all should be well. (Of course, near the
9037 start of the subject, fewer characters may be present; in that case all
9038 characters should be retained.)
9040 From release 8.33, there is a more accurate way of deciding which char-
9041 acters to retain. Instead of subtracting the length of the longest
9042 lookbehind from the earliest inspected character (offsets[0]), the
9043 match start position (offsets[2]) should be used, and the next match
9044 attempt started at the offsets[2] character by setting the startoffset
9045 argument of pcre_exec() or pcre_dfa_exec().
9047 For example, if the pattern "(?<=123)abc" is partially matched against
9048 the string "xx123a", the three offset values returned are 2, 6, and 5.
9049 This indicates that the matching process that gave a partial match
9050 started at offset 5, but the characters "123a" were all inspected. The
9051 maximum lookbehind for that pattern is 3, so taking that away from 5
9052 shows that we need only keep "123a", and the next match attempt can be
9053 started at offset 3 (that is, at "a") when further characters have been
9054 added. When the match start is not the earliest inspected character,
9055 pcretest shows it explicitly:
9059 Partial match at offset 5: 123a
9061 3. Because a partial match must always contain at least one character,
9062 what might be considered a partial match of an empty string actually
9063 gives a "no match" result. For example:
9069 If the next segment begins "cx", a match should be found, but this will
9070 only happen if characters from the previous segment are retained. For
9071 this reason, a "no match" result should be interpreted as "partial
9072 match of an empty string" when the pattern contains lookbehinds.
9074 4. Matching a subject string that is split into multiple segments may
9075 not always produce exactly the same result as matching over one single
9076 long string, especially when PCRE_PARTIAL_SOFT is used. The section
9077 "Partial Matching and Word Boundaries" above describes an issue that
9078 arises if the pattern ends with \b or \B. Another kind of difference
9079 may occur when there are multiple matching possibilities, because (for
9080 PCRE_PARTIAL_SOFT) a partial match result is given only when there are
9081 no completed matches. This means that as soon as the shortest match has
9082 been found, continuation to a new subject segment is no longer possi-
9083 ble. Consider again this pcretest example:
9096 The first data line passes the string "dogsb" to a standard matching
9097 function, setting the PCRE_PARTIAL_SOFT option. Although the string is
9098 a partial match for "dogsbody", the result is not PCRE_ERROR_PARTIAL,
9099 because the shorter string "dog" is a complete match. Similarly, when
9100 the subject is presented to a DFA matching function in several parts
9101 ("do" and "gsb" being the first two) the match stops when "dog" has
9102 been found, and it is not possible to continue. On the other hand, if
9103 "dogsbody" is presented as a single string, a DFA matching function
9106 Because of these problems, it is best to use PCRE_PARTIAL_HARD when
9107 matching multi-segment data. The example above then behaves differ-
9112 Partial match: dogsb
9118 5. Patterns that contain alternatives at the top level which do not all
9119 start with the same pattern item may not work as expected when
9120 PCRE_DFA_RESTART is used. For example, consider this pattern:
9124 If the first part of the subject is "ABC123", a partial match of the
9125 first alternative is found at offset 3. There is no partial match for
9126 the second alternative, because such a match does not start at the same
9127 point in the subject string. Attempting to continue with the string
9128 "7890" does not yield a match because only those alternatives that
9129 match at one point in the subject are remembered. The problem arises
9130 because the start of the second alternative matches within the first
9131 alternative. There is no problem with anchored patterns or patterns
9136 where no string can be a partial match for both alternatives. This is
9137 not a problem if a standard matching function is used, because the
9138 entire match has to be rerun each time:
9146 Of course, instead of using PCRE_DFA_RESTART, the same technique of re-
9147 running the entire match can also be used with the DFA matching func-
9148 tions. Another possibility is to work with two buffers. If a partial
9149 match at offset n in the first buffer is followed by "no match" when
9150 PCRE_DFA_RESTART is used on the second buffer, you can then try a new
9151 match starting at offset n+1 in the first buffer.
9157 University Computing Service
9158 Cambridge CB2 3QH, England.
9163 Last updated: 02 July 2013
9164 Copyright (c) 1997-2013 University of Cambridge.
9165 ------------------------------------------------------------------------------
9168 PCREPRECOMPILE(3) Library Functions Manual PCREPRECOMPILE(3)
9173 PCRE - Perl-compatible regular expressions
9175 SAVING AND RE-USING PRECOMPILED PCRE PATTERNS
9177 If you are running an application that uses a large number of regular
9178 expression patterns, it may be useful to store them in a precompiled
9179 form instead of having to compile them every time the application is
9180 run. If you are not using any private character tables (see the
9181 pcre_maketables() documentation), this is relatively straightforward.
9182 If you are using private tables, it is a little bit more complicated.
9183 However, if you are using the just-in-time optimization feature, it is
9184 not possible to save and reload the JIT data.
9186 If you save compiled patterns to a file, you can copy them to a differ-
9187 ent host and run them there. If the two hosts have different endianness
9188 (byte order), you should run the pcre[16|32]_pat-
9189 tern_to_host_byte_order() function on the new host before trying to
9190 match the pattern. The matching functions return PCRE_ERROR_BADENDIAN-
9191 NESS if they detect a pattern with the wrong endianness.
9193 Compiling regular expressions with one version of PCRE for use with a
9194 different version is not guaranteed to work and may cause crashes, and
9195 saving and restoring a compiled pattern loses any JIT optimization
9199 SAVING A COMPILED PATTERN
9201 The value returned by pcre[16|32]_compile() points to a single block of
9202 memory that holds the compiled pattern and associated data. You can
9203 find the length of this block in bytes by calling
9204 pcre[16|32]_fullinfo() with an argument of PCRE_INFO_SIZE. You can then
9205 save the data in any appropriate manner. Here is sample code for the
9206 8-bit library that compiles a pattern and writes it to a file. It
9207 assumes that the variable fd refers to a file that is open for output:
9209 int erroroffset, rc, size;
9213 re = pcre_compile("my pattern", 0, &error, &erroroffset, NULL);
9214 if (re == NULL) { ... handle errors ... }
9215 rc = pcre_fullinfo(re, NULL, PCRE_INFO_SIZE, &size);
9216 if (rc < 0) { ... handle errors ... }
9217 rc = fwrite(re, 1, size, fd);
9218 if (rc != size) { ... handle errors ... }
9220 In this example, the bytes that comprise the compiled pattern are
9221 copied exactly. Note that this is binary data that may contain any of
9222 the 256 possible byte values. On systems that make a distinction
9223 between binary and non-binary data, be sure that the file is opened for
9226 If you want to write more than one pattern to a file, you will have to
9227 devise a way of separating them. For binary data, preceding each pat-
9228 tern with its length is probably the most straightforward approach.
9229 Another possibility is to write out the data in hexadecimal instead of
9230 binary, one pattern to a line.
9232 Saving compiled patterns in a file is only one possible way of storing
9233 them for later use. They could equally well be saved in a database, or
9234 in the memory of some daemon process that passes them via sockets to
9235 the processes that want them.
9237 If the pattern has been studied, it is also possible to save the normal
9238 study data in a similar way to the compiled pattern itself. However, if
9239 the PCRE_STUDY_JIT_COMPILE was used, the just-in-time data that is cre-
9240 ated cannot be saved because it is too dependent on the current envi-
9241 ronment. When studying generates additional information,
9242 pcre[16|32]_study() returns a pointer to a pcre[16|32]_extra data
9243 block. Its format is defined in the section on matching a pattern in
9244 the pcreapi documentation. The study_data field points to the binary
9245 study data, and this is what you must save (not the pcre[16|32]_extra
9246 block itself). The length of the study data can be obtained by calling
9247 pcre[16|32]_fullinfo() with an argument of PCRE_INFO_STUDYSIZE. Remem-
9248 ber to check that pcre[16|32]_study() did return a non-NULL value
9249 before trying to save the study data.
9252 RE-USING A PRECOMPILED PATTERN
9254 Re-using a precompiled pattern is straightforward. Having reloaded it
9255 into main memory, called pcre[16|32]_pattern_to_host_byte_order() if
9256 necessary, you pass its pointer to pcre[16|32]_exec() or
9257 pcre[16|32]_dfa_exec() in the usual way.
9259 However, if you passed a pointer to custom character tables when the
9260 pattern was compiled (the tableptr argument of pcre[16|32]_compile()),
9261 you must now pass a similar pointer to pcre[16|32]_exec() or
9262 pcre[16|32]_dfa_exec(), because the value saved with the compiled pat-
9263 tern will obviously be nonsense. A field in a pcre[16|32]_extra() block
9264 is used to pass this data, as described in the section on matching a
9265 pattern in the pcreapi documentation.
9267 Warning: The tables that pcre_exec() and pcre_dfa_exec() use must be
9268 the same as those that were used when the pattern was compiled. If this
9269 is not the case, the behaviour is undefined.
9271 If you did not provide custom character tables when the pattern was
9272 compiled, the pointer in the compiled pattern is NULL, which causes the
9273 matching functions to use PCRE's internal tables. Thus, you do not need
9274 to take any special action at run time in this case.
9276 If you saved study data with the compiled pattern, you need to create
9277 your own pcre[16|32]_extra data block and set the study_data field to
9278 point to the reloaded study data. You must also set the
9279 PCRE_EXTRA_STUDY_DATA bit in the flags field to indicate that study
9280 data is present. Then pass the pcre[16|32]_extra block to the matching
9281 function in the usual way. If the pattern was studied for just-in-time
9282 optimization, that data cannot be saved, and so is lost by a
9286 COMPATIBILITY WITH DIFFERENT PCRE RELEASES
9288 In general, it is safest to recompile all saved patterns when you
9289 update to a new PCRE release, though not all updates actually require
9296 University Computing Service
9297 Cambridge CB2 3QH, England.
9302 Last updated: 12 November 2013
9303 Copyright (c) 1997-2013 University of Cambridge.
9304 ------------------------------------------------------------------------------
9307 PCREPERFORM(3) Library Functions Manual PCREPERFORM(3)
9312 PCRE - Perl-compatible regular expressions
9316 Two aspects of performance are discussed below: memory usage and pro-
9317 cessing time. The way you express your pattern as a regular expression
9318 can affect both of them.
9321 COMPILED PATTERN MEMORY USAGE
9323 Patterns are compiled by PCRE into a reasonably efficient interpretive
9324 code, so that most simple patterns do not use much memory. However,
9325 there is one case where the memory usage of a compiled pattern can be
9326 unexpectedly large. If a parenthesized subpattern has a quantifier with
9327 a minimum greater than 1 and/or a limited maximum, the whole subpattern
9328 is repeated in the compiled code. For example, the pattern
9332 is compiled as if it were
9334 (abc|def)(abc|def)((abc|def)(abc|def)?)?
9336 (Technical aside: It is done this way so that backtrack points within
9337 each of the repetitions can be independently maintained.)
9339 For regular expressions whose quantifiers use only small numbers, this
9340 is not usually a problem. However, if the numbers are large, and par-
9341 ticularly if such repetitions are nested, the memory usage can become
9342 an embarrassment. For example, the very simple pattern
9344 ((ab){1,1000}c){1,3}
9346 uses 51K bytes when compiled using the 8-bit library. When PCRE is com-
9347 piled with its default internal pointer size of two bytes, the size
9348 limit on a compiled pattern is 64K data units, and this is reached with
9349 the above pattern if the outer repetition is increased from 3 to 4.
9350 PCRE can be compiled to use larger internal pointers and thus handle
9351 larger compiled patterns, but it is better to try to rewrite your pat-
9352 tern to use less memory if you can.
9354 One way of reducing the memory usage for such patterns is to make use
9355 of PCRE's "subroutine" facility. Re-writing the above pattern as
9357 ((ab)(?2){0,999}c)(?1){0,2}
9359 reduces the memory requirements to 18K, and indeed it remains under 20K
9360 even with the outer repetition increased to 100. However, this pattern
9361 is not exactly equivalent, because the "subroutine" calls are treated
9362 as atomic groups into which there can be no backtracking if there is a
9363 subsequent matching failure. Therefore, PCRE cannot do this kind of
9364 rewriting automatically. Furthermore, there is a noticeable loss of
9365 speed when executing the modified pattern. Nevertheless, if the atomic
9366 grouping is not a problem and the loss of speed is acceptable, this
9367 kind of rewriting will allow you to process patterns that PCRE cannot
9371 STACK USAGE AT RUN TIME
9373 When pcre_exec() or pcre[16|32]_exec() is used for matching, certain
9374 kinds of pattern can cause it to use large amounts of the process
9375 stack. In some environments the default process stack is quite small,
9376 and if it runs out the result is often SIGSEGV. This issue is probably
9377 the most frequently raised problem with PCRE. Rewriting your pattern
9378 can often help. The pcrestack documentation discusses this issue in
9384 Certain items in regular expression patterns are processed more effi-
9385 ciently than others. It is more efficient to use a character class like
9386 [aeiou] than a set of single-character alternatives such as
9387 (a|e|i|o|u). In general, the simplest construction that provides the
9388 required behaviour is usually the most efficient. Jeffrey Friedl's book
9389 contains a lot of useful general discussion about optimizing regular
9390 expressions for efficient performance. This document contains a few
9391 observations about PCRE.
9393 Using Unicode character properties (the \p, \P, and \X escapes) is
9394 slow, because PCRE has to use a multi-stage table lookup whenever it
9395 needs a character's property. If you can find an alternative pattern
9396 that does not use character properties, it will probably be faster.
9398 By default, the escape sequences \b, \d, \s, and \w, and the POSIX
9399 character classes such as [:alpha:] do not use Unicode properties,
9400 partly for backwards compatibility, and partly for performance reasons.
9401 However, you can set PCRE_UCP if you want Unicode character properties
9402 to be used. This can double the matching time for items such as \d,
9403 when matched with a traditional matching function; the performance loss
9404 is less with a DFA matching function, and in both cases there is not
9405 much difference for \b.
9407 When a pattern begins with .* not in parentheses, or in parentheses
9408 that are not the subject of a backreference, and the PCRE_DOTALL option
9409 is set, the pattern is implicitly anchored by PCRE, since it can match
9410 only at the start of a subject string. However, if PCRE_DOTALL is not
9411 set, PCRE cannot make this optimization, because the . metacharacter
9412 does not then match a newline, and if the subject string contains new-
9413 lines, the pattern may match from the character immediately following
9414 one of them instead of from the very start. For example, the pattern
9418 matches the subject "first\nand second" (where \n stands for a newline
9419 character), with the match starting at the seventh character. In order
9420 to do this, PCRE has to retry the match starting after every newline in
9423 If you are using such a pattern with subject strings that do not con-
9424 tain newlines, the best performance is obtained by setting PCRE_DOTALL,
9425 or starting the pattern with ^.* or ^.*? to indicate explicit anchor-
9426 ing. That saves PCRE from having to scan along the subject looking for
9427 a newline to restart at.
9429 Beware of patterns that contain nested indefinite repeats. These can
9430 take a long time to run when applied to a string that does not match.
9431 Consider the pattern fragment
9435 This can match "aaaa" in 16 different ways, and this number increases
9436 very rapidly as the string gets longer. (The * repeat can match 0, 1,
9437 2, 3, or 4 times, and for each of those cases other than 0 or 4, the +
9438 repeats can match different numbers of times.) When the remainder of
9439 the pattern is such that the entire match is going to fail, PCRE has in
9440 principle to try every possible variation, and this can take an
9441 extremely long time, even for relatively short strings.
9443 An optimization catches some of the more simple cases such as
9447 where a literal character follows. Before embarking on the standard
9448 matching procedure, PCRE checks that there is a "b" later in the sub-
9449 ject string, and if there is not, it fails the match immediately. How-
9450 ever, when there is no following literal this optimization cannot be
9451 used. You can see the difference by comparing the behaviour of
9455 with the pattern above. The former gives a failure almost instantly
9456 when applied to a whole line of "a" characters, whereas the latter
9457 takes an appreciable time with strings longer than about 20 characters.
9459 In many cases, the solution to this kind of performance issue is to use
9460 an atomic group or a possessive quantifier.
9466 University Computing Service
9467 Cambridge CB2 3QH, England.
9472 Last updated: 25 August 2012
9473 Copyright (c) 1997-2012 University of Cambridge.
9474 ------------------------------------------------------------------------------
9477 PCREPOSIX(3) Library Functions Manual PCREPOSIX(3)
9482 PCRE - Perl-compatible regular expressions.
9486 #include <pcreposix.h>
9488 int regcomp(regex_t *preg, const char *pattern,
9491 int regexec(regex_t *preg, const char *string,
9492 size_t nmatch, regmatch_t pmatch[], int eflags);
9493 size_t regerror(int errcode, const regex_t *preg,
9494 char *errbuf, size_t errbuf_size);
9496 void regfree(regex_t *preg);
9501 This set of functions provides a POSIX-style API for the PCRE regular
9502 expression 8-bit library. See the pcreapi documentation for a descrip-
9503 tion of PCRE's native API, which contains much additional functional-
9504 ity. There is no POSIX-style wrapper for PCRE's 16-bit and 32-bit
9507 The functions described here are just wrapper functions that ultimately
9508 call the PCRE native API. Their prototypes are defined in the
9509 pcreposix.h header file, and on Unix systems the library itself is
9510 called pcreposix.a, so can be accessed by adding -lpcreposix to the
9511 command for linking an application that uses them. Because the POSIX
9512 functions call the native ones, it is also necessary to add -lpcre.
9514 I have implemented only those POSIX option bits that can be reasonably
9515 mapped to PCRE native options. In addition, the option REG_EXTENDED is
9516 defined with the value zero. This has no effect, but since programs
9517 that are written to the POSIX interface often use it, this makes it
9518 easier to slot in PCRE as a replacement library. Other POSIX options
9519 are not even defined.
9521 There are also some other options that are not defined by POSIX. These
9522 have been added at the request of users who want to make use of certain
9523 PCRE-specific features via the POSIX calling interface.
9525 When PCRE is called via these functions, it is only the API that is
9526 POSIX-like in style. The syntax and semantics of the regular expres-
9527 sions themselves are still those of Perl, subject to the setting of
9528 various PCRE options, as described below. "POSIX-like in style" means
9529 that the API approximates to the POSIX definition; it is not fully
9530 POSIX-compatible, and in multi-byte encoding domains it is probably
9531 even less compatible.
9533 The header for these functions is supplied as pcreposix.h to avoid any
9534 potential clash with other POSIX libraries. It can, of course, be
9535 renamed or aliased as regex.h, which is the "correct" name. It provides
9536 two structure types, regex_t for compiled internal forms, and reg-
9537 match_t for returning captured substrings. It also defines some con-
9538 stants whose names start with "REG_"; these are used for setting
9539 options and identifying error codes.
9544 The function regcomp() is called to compile a pattern into an internal
9545 form. The pattern is a C string terminated by a binary zero, and is
9546 passed in the argument pattern. The preg argument is a pointer to a
9547 regex_t structure that is used as a base for storing information about
9548 the compiled regular expression.
9550 The argument cflags is either zero, or contains one or more of the bits
9551 defined by the following macros:
9555 The PCRE_DOTALL option is set when the regular expression is passed for
9556 compilation to the native function. Note that REG_DOTALL is not part of
9561 The PCRE_CASELESS option is set when the regular expression is passed
9562 for compilation to the native function.
9566 The PCRE_MULTILINE option is set when the regular expression is passed
9567 for compilation to the native function. Note that this does not mimic
9568 the defined POSIX behaviour for REG_NEWLINE (see the following sec-
9573 The PCRE_NO_AUTO_CAPTURE option is set when the regular expression is
9574 passed for compilation to the native function. In addition, when a pat-
9575 tern that is compiled with this flag is passed to regexec() for match-
9576 ing, the nmatch and pmatch arguments are ignored, and no captured
9577 strings are returned.
9581 The PCRE_UCP option is set when the regular expression is passed for
9582 compilation to the native function. This causes PCRE to use Unicode
9583 properties when matchine \d, \w, etc., instead of just recognizing
9584 ASCII values. Note that REG_UTF8 is not part of the POSIX standard.
9588 The PCRE_UNGREEDY option is set when the regular expression is passed
9589 for compilation to the native function. Note that REG_UNGREEDY is not
9590 part of the POSIX standard.
9594 The PCRE_UTF8 option is set when the regular expression is passed for
9595 compilation to the native function. This causes the pattern itself and
9596 all data strings used for matching it to be treated as UTF-8 strings.
9597 Note that REG_UTF8 is not part of the POSIX standard.
9599 In the absence of these flags, no options are passed to the native
9600 function. This means the the regex is compiled with PCRE default
9601 semantics. In particular, the way it handles newline characters in the
9602 subject string is the Perl way, not the POSIX way. Note that setting
9603 PCRE_MULTILINE has only some of the effects specified for REG_NEWLINE.
9604 It does not affect the way newlines are matched by . (they are not) or
9605 by a negative class such as [^a] (they are).
9607 The yield of regcomp() is zero on success, and non-zero otherwise. The
9608 preg structure is filled in on success, and one member of the structure
9609 is public: re_nsub contains the number of capturing subpatterns in the
9610 regular expression. Various error codes are defined in the header file.
9612 NOTE: If the yield of regcomp() is non-zero, you must not attempt to
9613 use the contents of the preg structure. If, for example, you pass it to
9614 regexec(), the result is undefined and your program is likely to crash.
9617 MATCHING NEWLINE CHARACTERS
9619 This area is not simple, because POSIX and Perl take different views of
9620 things. It is not possible to get PCRE to obey POSIX semantics, but
9621 then PCRE was never intended to be a POSIX engine. The following table
9622 lists the different possibilities for matching newline characters in
9627 . matches newline no PCRE_DOTALL
9628 newline matches [^a] yes not changeable
9629 $ matches \n at end yes PCRE_DOLLARENDONLY
9630 $ matches \n in middle no PCRE_MULTILINE
9631 ^ matches \n in middle no PCRE_MULTILINE
9633 This is the equivalent table for POSIX:
9637 . matches newline yes REG_NEWLINE
9638 newline matches [^a] yes REG_NEWLINE
9639 $ matches \n at end no REG_NEWLINE
9640 $ matches \n in middle no REG_NEWLINE
9641 ^ matches \n in middle no REG_NEWLINE
9643 PCRE's behaviour is the same as Perl's, except that there is no equiva-
9644 lent for PCRE_DOLLAR_ENDONLY in Perl. In both PCRE and Perl, there is
9645 no way to stop newline from matching [^a].
9647 The default POSIX newline handling can be obtained by setting
9648 PCRE_DOTALL and PCRE_DOLLAR_ENDONLY, but there is no way to make PCRE
9649 behave exactly as for the REG_NEWLINE action.
9654 The function regexec() is called to match a compiled pattern preg
9655 against a given string, which is by default terminated by a zero byte
9656 (but see REG_STARTEND below), subject to the options in eflags. These
9661 The PCRE_NOTBOL option is set when calling the underlying PCRE matching
9666 The PCRE_NOTEMPTY option is set when calling the underlying PCRE match-
9667 ing function. Note that REG_NOTEMPTY is not part of the POSIX standard.
9668 However, setting this option can give more POSIX-like behaviour in some
9673 The PCRE_NOTEOL option is set when calling the underlying PCRE matching
9678 The string is considered to start at string + pmatch[0].rm_so and to
9679 have a terminating NUL located at string + pmatch[0].rm_eo (there need
9680 not actually be a NUL at that location), regardless of the value of
9681 nmatch. This is a BSD extension, compatible with but not specified by
9682 IEEE Standard 1003.2 (POSIX.2), and should be used with caution in
9683 software intended to be portable to other systems. Note that a non-zero
9684 rm_so does not imply REG_NOTBOL; REG_STARTEND affects only the location
9685 of the string, not how it is matched.
9687 If the pattern was compiled with the REG_NOSUB flag, no data about any
9688 matched strings is returned. The nmatch and pmatch arguments of
9689 regexec() are ignored.
9691 If the value of nmatch is zero, or if the value pmatch is NULL, no data
9692 about any matched strings is returned.
9694 Otherwise,the portion of the string that was matched, and also any cap-
9695 tured substrings, are returned via the pmatch argument, which points to
9696 an array of nmatch structures of type regmatch_t, containing the mem-
9697 bers rm_so and rm_eo. These contain the offset to the first character
9698 of each substring and the offset to the first character after the end
9699 of each substring, respectively. The 0th element of the vector relates
9700 to the entire portion of string that was matched; subsequent elements
9701 relate to the capturing subpatterns of the regular expression. Unused
9702 entries in the array have both structure members set to -1.
9704 A successful match yields a zero return; various error codes are
9705 defined in the header file, of which REG_NOMATCH is the "expected"
9711 The regerror() function maps a non-zero errorcode from either regcomp()
9712 or regexec() to a printable message. If preg is not NULL, the error
9713 should have arisen from the use of that structure. A message terminated
9714 by a binary zero is placed in errbuf. The length of the message,
9715 including the zero, is limited to errbuf_size. The yield of the func-
9716 tion is the size of buffer needed to hold the whole message.
9721 Compiling a regular expression causes memory to be allocated and asso-
9722 ciated with the preg structure. The function regfree() frees all such
9723 memory, after which preg may no longer be used as a compiled expres-
9730 University Computing Service
9731 Cambridge CB2 3QH, England.
9736 Last updated: 09 January 2012
9737 Copyright (c) 1997-2012 University of Cambridge.
9738 ------------------------------------------------------------------------------
9741 PCRECPP(3) Library Functions Manual PCRECPP(3)
9746 PCRE - Perl-compatible regular expressions.
9748 SYNOPSIS OF C++ WRAPPER
9750 #include <pcrecpp.h>
9755 The C++ wrapper for PCRE was provided by Google Inc. Some additional
9756 functionality was added by Giuseppe Maxia. This brief man page was con-
9757 structed from the notes in the pcrecpp.h file, which should be con-
9758 sulted for further details. Note that the C++ wrapper supports only the
9759 original 8-bit PCRE library. There is no 16-bit or 32-bit support at
9765 The "FullMatch" operation checks that supplied text matches a supplied
9766 pattern exactly. If pointer arguments are supplied, it copies matched
9767 sub-strings that match sub-patterns into them.
9769 Example: successful match
9770 pcrecpp::RE re("h.*o");
9771 re.FullMatch("hello");
9773 Example: unsuccessful match (requires full match):
9774 pcrecpp::RE re("e");
9775 !re.FullMatch("hello");
9777 Example: creating a temporary RE object:
9778 pcrecpp::RE("h.*o").FullMatch("hello");
9780 You can pass in a "const char*" or a "string" for "text". The examples
9781 below tend to use a const char*. You can, as in the different examples
9782 above, store the RE object explicitly in a variable or use a temporary
9783 RE object. The examples below use one mode or the other arbitrarily.
9784 Either could correctly be used for any of these examples.
9786 You must supply extra pointer arguments to extract matched subpieces.
9788 Example: extracts "ruby" into "s" and 1234 into "i"
9791 pcrecpp::RE re("(\\w+):(\\d+)");
9792 re.FullMatch("ruby:1234", &s, &i);
9794 Example: does not try to extract any extra sub-patterns
9795 re.FullMatch("ruby:1234", &s);
9797 Example: does not try to extract into NULL
9798 re.FullMatch("ruby:1234", NULL, &i);
9800 Example: integer overflow causes failure
9801 !re.FullMatch("ruby:1234567891234", NULL, &i);
9803 Example: fails because there aren't enough sub-patterns:
9804 !pcrecpp::RE("\\w+:\\d+").FullMatch("ruby:1234", &s);
9806 Example: fails because string cannot be stored in integer
9807 !pcrecpp::RE("(.*)").FullMatch("ruby", &i);
9809 The provided pointer arguments can be pointers to any scalar numeric
9812 string (matched piece is copied to string)
9813 StringPiece (StringPiece is mutated to point to matched piece)
9814 T (where "bool T::ParseFrom(const char*, int)" exists)
9815 NULL (the corresponding matched sub-pattern is not copied)
9817 The function returns true iff all of the following conditions are sat-
9820 a. "text" matches "pattern" exactly;
9822 b. The number of matched sub-patterns is >= number of supplied
9825 c. The "i"th argument has a suitable type for holding the
9826 string captured as the "i"th sub-pattern. If you pass in
9827 void * NULL for the "i"th argument, or a non-void * NULL
9828 of the correct type, or pass fewer arguments than the
9829 number of sub-patterns, "i"th captured sub-pattern is
9832 CAVEAT: An optional sub-pattern that does not exist in the matched
9833 string is assigned the empty string. Therefore, the following will
9834 return false (because the empty string is not a valid number):
9837 pcrecpp::RE::FullMatch("abc", "[a-z]+(\\d+)?", &number);
9839 The matching interface supports at most 16 arguments per call. If you
9840 need more, consider using the more general interface
9841 pcrecpp::RE::DoMatch. See pcrecpp.h for the signature for DoMatch.
9843 NOTE: Do not use no_arg, which is used internally to mark the end of a
9844 list of optional arguments, as a placeholder for missing arguments, as
9845 this can lead to segfaults.
9848 QUOTING METACHARACTERS
9850 You can use the "QuoteMeta" operation to insert backslashes before all
9851 potentially meaningful characters in a string. The returned string,
9852 used as a regular expression, will exactly match the original string.
9855 string quoted = RE::QuoteMeta(unquoted);
9857 Note that it's legal to escape a character even if it has no special
9858 meaning in a regular expression -- so this function does that. (This
9859 also makes it identical to the perl function of the same name; see
9860 "perldoc -f quotemeta".) For example, "1.5-2.0?" becomes
9866 You can use the "PartialMatch" operation when you want the pattern to
9867 match any substring of the text.
9869 Example: simple search for a string:
9870 pcrecpp::RE("ell").PartialMatch("hello");
9872 Example: find first number in a string:
9874 pcrecpp::RE re("(\\d+)");
9875 re.PartialMatch("x*100 + 20", &number);
9876 assert(number == 100);
9879 UTF-8 AND THE MATCHING INTERFACE
9881 By default, pattern and text are plain text, one byte per character.
9882 The UTF8 flag, passed to the constructor, causes both pattern and
9883 string to be treated as UTF-8 text, still a byte stream but potentially
9884 multiple bytes per character. In practice, the text is likelier to be
9885 UTF-8 than the pattern, but the match returned may depend on the UTF8
9886 flag, so always use it when matching UTF8 text. For example, "." will
9887 match one byte normally but with UTF8 set may match up to three bytes
9888 of a multi-byte character.
9891 pcrecpp::RE_Options options;
9893 pcrecpp::RE re(utf8_pattern, options);
9894 re.FullMatch(utf8_string);
9896 Example: using the convenience function UTF8():
9897 pcrecpp::RE re(utf8_pattern, pcrecpp::UTF8());
9898 re.FullMatch(utf8_string);
9900 NOTE: The UTF8 flag is ignored if pcre was not configured with the
9904 PASSING MODIFIERS TO THE REGULAR EXPRESSION ENGINE
9906 PCRE defines some modifiers to change the behavior of the regular
9907 expression engine. The C++ wrapper defines an auxiliary class,
9908 RE_Options, as a vehicle to pass such modifiers to a RE class. Cur-
9909 rently, the following modifiers are supported:
9911 modifier description Perl corresponding
9913 PCRE_CASELESS case insensitive match /i
9914 PCRE_MULTILINE multiple lines match /m
9915 PCRE_DOTALL dot matches newlines /s
9916 PCRE_DOLLAR_ENDONLY $ matches only at end N/A
9917 PCRE_EXTRA strict escape parsing N/A
9918 PCRE_EXTENDED ignore white spaces /x
9919 PCRE_UTF8 handles UTF8 chars built-in
9920 PCRE_UNGREEDY reverses * and *? N/A
9921 PCRE_NO_AUTO_CAPTURE disables capturing parens N/A (*)
9923 (*) Both Perl and PCRE allow non capturing parentheses by means of the
9924 "?:" modifier within the pattern itself. e.g. (?:ab|cd) does not cap-
9925 ture, while (ab|cd) does.
9927 For a full account on how each modifier works, please check the PCRE
9930 For each modifier, there are two member functions whose name is made
9931 out of the modifier in lowercase, without the "PCRE_" prefix. For
9932 instance, PCRE_CASELESS is handled by
9936 which returns true if the modifier is set, and
9938 RE_Options & set_caseless(bool)
9940 which sets or unsets the modifier. Moreover, PCRE_EXTRA_MATCH_LIMIT can
9941 be accessed through the set_match_limit() and match_limit() member
9942 functions. Setting match_limit to a non-zero value will limit the exe-
9943 cution of pcre to keep it from doing bad things like blowing the stack
9944 or taking an eternity to return a result. A value of 5000 is good
9945 enough to stop stack blowup in a 2MB thread stack. Setting match_limit
9946 to zero disables match limiting. Alternatively, you can call
9947 match_limit_recursion() which uses PCRE_EXTRA_MATCH_LIMIT_RECURSION to
9948 limit how much PCRE recurses. match_limit() limits the number of
9949 matches PCRE does; match_limit_recursion() limits the depth of internal
9950 recursion, and therefore the amount of stack that is used.
9952 Normally, to pass one or more modifiers to a RE class, you declare a
9953 RE_Options object, set the appropriate options, and pass this object to
9954 a RE constructor. Example:
9957 opt.set_caseless(true);
9958 if (RE("HELLO", opt).PartialMatch("hello world")) ...
9960 RE_options has two constructors. The default constructor takes no argu-
9961 ments and creates a set of flags that are off by default. The optional
9962 parameter option_flags is to facilitate transfer of legacy code from C
9963 programs. This lets you do
9966 RE_Options(PCRE_CASELESS|PCRE_MULTILINE)).PartialMatch(str);
9968 However, new code is better off doing
9971 RE_Options().set_caseless(true).set_multiline(true))
9974 If you are going to pass one of the most used modifiers, there are some
9975 convenience functions that return a RE_Options class with the appropri-
9976 ate modifier already set: CASELESS(), UTF8(), MULTILINE(), DOTALL(),
9979 If you need to set several options at once, and you don't want to go
9980 through the pains of declaring a RE_Options object and setting several
9981 options, there is a parallel method that give you such ability on the
9982 fly. You can concatenate several set_xxxxx() member functions, since
9983 each of them returns a reference to its class object. For example, to
9984 pass PCRE_CASELESS, PCRE_EXTENDED, and PCRE_MULTILINE to a RE with one
9985 statement, you may write:
9987 RE(" ^ xyz \\s+ .* blah$",
9991 .set_multiline(true)).PartialMatch(sometext);
9994 SCANNING TEXT INCREMENTALLY
9996 The "Consume" operation may be useful if you want to repeatedly match
9997 regular expressions at the front of a string and skip over them as they
9998 match. This requires use of the "StringPiece" type, which represents a
9999 sub-range of a real string. Like RE, StringPiece is defined in the
10002 Example: read lines of the form "var = value" from a string.
10003 string contents = ...; // Fill string somehow
10004 pcrecpp::StringPiece input(contents); // Wrap in a StringPiece
10008 pcrecpp::RE re("(\\w+) = (\\d+)\n");
10009 while (re.Consume(&input, &var, &value)) {
10013 Each successful call to "Consume" will set "var/value", and also
10014 advance "input" so it points past the matched text.
10016 The "FindAndConsume" operation is similar to "Consume" but does not
10017 anchor your match at the beginning of the string. For example, you
10018 could extract all words from a string by repeatedly calling
10020 pcrecpp::RE("(\\w+)").FindAndConsume(&input, &word)
10023 PARSING HEX/OCTAL/C-RADIX NUMBERS
10025 By default, if you pass a pointer to a numeric value, the corresponding
10026 text is interpreted as a base-10 number. You can instead wrap the
10027 pointer with a call to one of the operators Hex(), Octal(), or CRadix()
10028 to interpret the text in another base. The CRadix operator interprets
10029 C-style "0" (base-8) and "0x" (base-16) prefixes, but defaults to
10034 pcrecpp::RE re("(.*) (.*) (.*) (.*)");
10035 re.FullMatch("100 40 0100 0x40",
10036 pcrecpp::Octal(&a), pcrecpp::Hex(&b),
10037 pcrecpp::CRadix(&c), pcrecpp::CRadix(&d));
10039 will leave 64 in a, b, c, and d.
10042 REPLACING PARTS OF STRINGS
10044 You can replace the first match of "pattern" in "str" with "rewrite".
10045 Within "rewrite", backslash-escaped digits (\1 to \9) can be used to
10046 insert text matching corresponding parenthesized group from the pat-
10047 tern. \0 in "rewrite" refers to the entire matching text. For example:
10049 string s = "yabba dabba doo";
10050 pcrecpp::RE("b+").Replace("d", &s);
10052 will leave "s" containing "yada dabba doo". The result is true if the
10053 pattern matches and a replacement occurs, false otherwise.
10055 GlobalReplace is like Replace except that it replaces all occurrences
10056 of the pattern in the string with the rewrite. Replacements are not
10057 subject to re-matching. For example:
10059 string s = "yabba dabba doo";
10060 pcrecpp::RE("b+").GlobalReplace("d", &s);
10062 will leave "s" containing "yada dada doo". It returns the number of
10065 Extract is like Replace, except that if the pattern matches, "rewrite"
10066 is copied into "out" (an additional argument) with substitutions. The
10067 non-matching portions of "text" are ignored. Returns true iff a match
10068 occurred and the extraction happened successfully; if no match occurs,
10069 the string is left unaffected.
10074 The C++ wrapper was contributed by Google Inc.
10075 Copyright (c) 2007 Google Inc.
10080 Last updated: 08 January 2012
10081 ------------------------------------------------------------------------------
10084 PCRESAMPLE(3) Library Functions Manual PCRESAMPLE(3)
10089 PCRE - Perl-compatible regular expressions
10091 PCRE SAMPLE PROGRAM
10093 A simple, complete demonstration program, to get you started with using
10094 PCRE, is supplied in the file pcredemo.c in the PCRE distribution. A
10095 listing of this program is given in the pcredemo documentation. If you
10096 do not have a copy of the PCRE distribution, you can save this listing
10097 to re-create pcredemo.c.
10099 The demonstration program, which uses the original PCRE 8-bit library,
10100 compiles the regular expression that is its first argument, and matches
10101 it against the subject string in its second argument. No PCRE options
10102 are set, and default character tables are used. If matching succeeds,
10103 the program outputs the portion of the subject that matched, together
10104 with the contents of any captured substrings.
10106 If the -g option is given on the command line, the program then goes on
10107 to check for further matches of the same regular expression in the same
10108 subject string. The logic is a little bit tricky because of the possi-
10109 bility of matching an empty string. Comments in the code explain what
10112 If PCRE is installed in the standard include and library directories
10113 for your operating system, you should be able to compile the demonstra-
10114 tion program using this command:
10116 gcc -o pcredemo pcredemo.c -lpcre
10118 If PCRE is installed elsewhere, you may need to add additional options
10119 to the command line. For example, on a Unix-like system that has PCRE
10120 installed in /usr/local, you can compile the demonstration program
10121 using a command like this:
10123 gcc -o pcredemo -I/usr/local/include pcredemo.c \
10124 -L/usr/local/lib -lpcre
10126 In a Windows environment, if you want to statically link the program
10127 against a non-dll pcre.a file, you must uncomment the line that defines
10128 PCRE_STATIC before including pcre.h, because otherwise the pcre_mal-
10129 loc() and pcre_free() exported functions will be declared
10130 __declspec(dllimport), with unwanted results.
10132 Once you have compiled and linked the demonstration program, you can
10133 run simple tests like this:
10135 ./pcredemo 'cat|dog' 'the cat sat on the mat'
10136 ./pcredemo -g 'cat|dog' 'the dog sat on the cat'
10138 Note that there is a much more comprehensive test program, called
10139 pcretest, which supports many more facilities for testing regular
10140 expressions and both PCRE libraries. The pcredemo program is provided
10141 as a simple coding example.
10143 If you try to run pcredemo when PCRE is not installed in the standard
10144 library directory, you may get an error like this on some operating
10145 systems (e.g. Solaris):
10147 ld.so.1: a.out: fatal: libpcre.so.0: open failed: No such file or
10150 This is caused by the way shared library support works on those sys-
10151 tems. You need to add
10155 (for example) to the compile command to get round this problem.
10161 University Computing Service
10162 Cambridge CB2 3QH, England.
10167 Last updated: 10 January 2012
10168 Copyright (c) 1997-2012 University of Cambridge.
10169 ------------------------------------------------------------------------------
10170 PCRELIMITS(3) Library Functions Manual PCRELIMITS(3)
10175 PCRE - Perl-compatible regular expressions
10177 SIZE AND OTHER LIMITATIONS
10179 There are some size limitations in PCRE but it is hoped that they will
10180 never in practice be relevant.
10182 The maximum length of a compiled pattern is approximately 64K data
10183 units (bytes for the 8-bit library, 16-bit units for the 16-bit
10184 library, and 32-bit units for the 32-bit library) if PCRE is compiled
10185 with the default internal linkage size, which is 2 bytes for the 8-bit
10186 and 16-bit libraries, and 4 bytes for the 32-bit library. If you want
10187 to process regular expressions that are truly enormous, you can compile
10188 PCRE with an internal linkage size of 3 or 4 (when building the 16-bit
10189 or 32-bit library, 3 is rounded up to 4). See the README file in the
10190 source distribution and the pcrebuild documentation for details. In
10191 these cases the limit is substantially larger. However, the speed of
10192 execution is slower.
10194 All values in repeating quantifiers must be less than 65536.
10196 There is no limit to the number of parenthesized subpatterns, but there
10197 can be no more than 65535 capturing subpatterns. There is, however, a
10198 limit to the depth of nesting of parenthesized subpatterns of all
10199 kinds. This is imposed in order to limit the amount of system stack
10200 used at compile time. The limit can be specified when PCRE is built;
10201 the default is 250.
10203 There is a limit to the number of forward references to subsequent sub-
10204 patterns of around 200,000. Repeated forward references with fixed
10205 upper limits, for example, (?2){0,100} when subpattern number 2 is to
10206 the right, are included in the count. There is no limit to the number
10207 of backward references.
10209 The maximum length of name for a named subpattern is 32 characters, and
10210 the maximum number of named subpatterns is 10000.
10212 The maximum length of a name in a (*MARK), (*PRUNE), (*SKIP), or
10213 (*THEN) verb is 255 for the 8-bit library and 65535 for the 16-bit and
10216 The maximum length of a subject string is the largest positive number
10217 that an integer variable can hold. However, when using the traditional
10218 matching function, PCRE uses recursion to handle subpatterns and indef-
10219 inite repetition. This means that the available stack space may limit
10220 the size of a subject string that can be processed by certain patterns.
10221 For a discussion of stack issues, see the pcrestack documentation.
10227 University Computing Service
10228 Cambridge CB2 3QH, England.
10233 Last updated: 05 November 2013
10234 Copyright (c) 1997-2013 University of Cambridge.
10235 ------------------------------------------------------------------------------
10238 PCRESTACK(3) Library Functions Manual PCRESTACK(3)
10243 PCRE - Perl-compatible regular expressions
10245 PCRE DISCUSSION OF STACK USAGE
10247 When you call pcre[16|32]_exec(), it makes use of an internal function
10248 called match(). This calls itself recursively at branch points in the
10249 pattern, in order to remember the state of the match so that it can
10250 back up and try a different alternative if the first one fails. As
10251 matching proceeds deeper and deeper into the tree of possibilities, the
10252 recursion depth increases. The match() function is also called in other
10253 circumstances, for example, whenever a parenthesized sub-pattern is
10254 entered, and in certain cases of repetition.
10256 Not all calls of match() increase the recursion depth; for an item such
10257 as a* it may be called several times at the same level, after matching
10258 different numbers of a's. Furthermore, in a number of cases where the
10259 result of the recursive call would immediately be passed back as the
10260 result of the current call (a "tail recursion"), the function is just
10263 The above comments apply when pcre[16|32]_exec() is run in its normal
10264 interpretive manner. If the pattern was studied with the
10265 PCRE_STUDY_JIT_COMPILE option, and just-in-time compiling was success-
10266 ful, and the options passed to pcre[16|32]_exec() were not incompati-
10267 ble, the matching process uses the JIT-compiled code instead of the
10268 match() function. In this case, the memory requirements are handled
10269 entirely differently. See the pcrejit documentation for details.
10271 The pcre[16|32]_dfa_exec() function operates in an entirely different
10272 way, and uses recursion only when there is a regular expression recur-
10273 sion or subroutine call in the pattern. This includes the processing of
10274 assertion and "once-only" subpatterns, which are handled like subrou-
10275 tine calls. Normally, these are never very deep, and the limit on the
10276 complexity of pcre[16|32]_dfa_exec() is controlled by the amount of
10277 workspace it is given. However, it is possible to write patterns with
10278 runaway infinite recursions; such patterns will cause
10279 pcre[16|32]_dfa_exec() to run out of stack. At present, there is no
10280 protection against this.
10282 The comments that follow do NOT apply to pcre[16|32]_dfa_exec(); they
10283 are relevant only for pcre[16|32]_exec() without the JIT optimization.
10285 Reducing pcre[16|32]_exec()'s stack usage
10287 Each time that match() is actually called recursively, it uses memory
10288 from the process stack. For certain kinds of pattern and data, very
10289 large amounts of stack may be needed, despite the recognition of "tail
10290 recursion". You can often reduce the amount of recursion, and there-
10291 fore the amount of stack used, by modifying the pattern that is being
10292 matched. Consider, for example, this pattern:
10296 It matches from wherever it starts until it encounters "<inet" or the
10297 end of the data, and is the kind of pattern that might be used when
10298 processing an XML file. Each iteration of the outer parentheses matches
10299 either one character that is not "<" or a "<" that is not followed by
10300 "inet". However, each time a parenthesis is processed, a recursion
10301 occurs, so this formulation uses a stack frame for each matched charac-
10302 ter. For a long string, a lot of stack is required. Consider now this
10303 rewritten pattern, which matches exactly the same strings:
10305 ([^<]++|<(?!inet))+
10307 This uses very much less stack, because runs of characters that do not
10308 contain "<" are "swallowed" in one item inside the parentheses. Recur-
10309 sion happens only when a "<" character that is not followed by "inet"
10310 is encountered (and we assume this is relatively rare). A possessive
10311 quantifier is used to stop any backtracking into the runs of non-"<"
10312 characters, but that is not related to stack usage.
10314 This example shows that one way of avoiding stack problems when match-
10315 ing long subject strings is to write repeated parenthesized subpatterns
10316 to match more than one character whenever possible.
10318 Compiling PCRE to use heap instead of stack for pcre[16|32]_exec()
10320 In environments where stack memory is constrained, you might want to
10321 compile PCRE to use heap memory instead of stack for remembering back-
10322 up points when pcre[16|32]_exec() is running. This makes it run a lot
10323 more slowly, however. Details of how to do this are given in the pcre-
10324 build documentation. When built in this way, instead of using the
10325 stack, PCRE obtains and frees memory by calling the functions that are
10326 pointed to by the pcre[16|32]_stack_malloc and pcre[16|32]_stack_free
10327 variables. By default, these point to malloc() and free(), but you can
10328 replace the pointers to cause PCRE to use your own functions. Since the
10329 block sizes are always the same, and are always freed in reverse order,
10330 it may be possible to implement customized memory handlers that are
10331 more efficient than the standard functions.
10333 Limiting pcre[16|32]_exec()'s stack usage
10335 You can set limits on the number of times that match() is called, both
10336 in total and recursively. If a limit is exceeded, pcre[16|32]_exec()
10337 returns an error code. Setting suitable limits should prevent it from
10338 running out of stack. The default values of the limits are very large,
10339 and unlikely ever to operate. They can be changed when PCRE is built,
10340 and they can also be set when pcre[16|32]_exec() is called. For details
10341 of these interfaces, see the pcrebuild documentation and the section on
10342 extra data for pcre[16|32]_exec() in the pcreapi documentation.
10344 As a very rough rule of thumb, you should reckon on about 500 bytes per
10345 recursion. Thus, if you want to limit your stack usage to 8Mb, you
10346 should set the limit at 16000 recursions. A 64Mb stack, on the other
10347 hand, can support around 128000 recursions.
10349 In Unix-like environments, the pcretest test program has a command line
10350 option (-S) that can be used to increase the size of its stack. As long
10351 as the stack is large enough, another option (-M) can be used to find
10352 the smallest limits that allow a particular pattern to match a given
10353 subject string. This is done by calling pcre[16|32]_exec() repeatedly
10354 with different limits.
10356 Obtaining an estimate of stack usage
10358 The actual amount of stack used per recursion can vary quite a lot,
10359 depending on the compiler that was used to build PCRE and the optimiza-
10360 tion or debugging options that were set for it. The rule of thumb value
10361 of 500 bytes mentioned above may be larger or smaller than what is
10362 actually needed. A better approximation can be obtained by running this
10367 The -C option causes pcretest to output information about the options
10368 with which PCRE was compiled. When -m is also given (before -C), infor-
10369 mation about stack use is given in a line like this:
10371 Match recursion uses stack: approximate frame size = 640 bytes
10373 The value is approximate because some recursions need a bit more (up to
10374 perhaps 16 more bytes).
10376 If the above command is given when PCRE is compiled to use the heap
10377 instead of the stack for recursion, the value that is output is the
10378 size of each block that is obtained from the heap.
10380 Changing stack size in Unix-like systems
10382 In Unix-like environments, there is not often a problem with the stack
10383 unless very long strings are involved, though the default limit on
10384 stack size varies from system to system. Values from 8Mb to 64Mb are
10385 common. You can find your default limit by running the command:
10389 Unfortunately, the effect of running out of stack is often SIGSEGV,
10390 though sometimes a more explicit error message is given. You can nor-
10391 mally increase the limit on stack size by code such as this:
10393 struct rlimit rlim;
10394 getrlimit(RLIMIT_STACK, &rlim);
10395 rlim.rlim_cur = 100*1024*1024;
10396 setrlimit(RLIMIT_STACK, &rlim);
10398 This reads the current limits (soft and hard) using getrlimit(), then
10399 attempts to increase the soft limit to 100Mb using setrlimit(). You
10400 must do this before calling pcre[16|32]_exec().
10402 Changing stack size in Mac OS X
10404 Using setrlimit(), as described above, should also work on Mac OS X. It
10405 is also possible to set a stack size when linking a program. There is a
10406 discussion about stack sizes in Mac OS X at this web site:
10407 http://developer.apple.com/qa/qa2005/qa1419.html.
10413 University Computing Service
10414 Cambridge CB2 3QH, England.
10419 Last updated: 24 June 2012
10420 Copyright (c) 1997-2012 University of Cambridge.
10421 ------------------------------------------------------------------------------